Personalized User Model LLP v. Google Inc.

Filing 727

NOTICE of Compliance by Google Inc. (Attachments: # 1 Exhibit A)(Palapura, Bindu)

EXHIBIT A

618
1

IN THE UNITED STATES DISTRICT COURT

2

IN AND FOR THE DISTRICT OF DELAWARE

3
4
5
6
7
8
9
10
11
12
13

- - PERSONALIZED USER MODEL, L.L.P.,
v

Plaintiff,

GOOGLE, INC.,

Defendant.
------------------------------------GOOGLE, INC.,
:
:
Counterclaimant,
:
v
:
:
PERSONALIZED USER MODEL, L.L.P.,
:
and YOCHAI KONIG,
:
:
Counterclaim-Defendants. :

16

19
20
21
22
23
24
25

NO. 09-525-LPS

Wilmington, Delaware
Wednesday, March 12, 2014
Jury Trial - Volume C

15

18

CIVIL ACTION

- - -

14

17

:
:
:
:
:
:
:
:

- - BEFORE:

HONORABLE LEONARD P. STARK, U.S.D.C.J.
- - -

APPEARANCES:

MORRIS NICHOLS ARSHT & TUNNELL, LLP
BY: KAREN JACOBS, ESQ.,
REGINA S.E. MURPHY, ESQ., and
JEREMY A. TIGAN, ESQ.
and
Valerie Gunning
Official Court Reporter

Brian P. Gaffigan
Official Court Reporter

619
1
2
3
4
5
6
7
8
9
10
11

APPEARANCES:

(Continued)
SNR DENTON, LLP
BY: MARK C. NELSON, ESQ.,
RICHARD D. SALGADO, ESQ., and
JUANITA DeLOACH, Ph.D., ESQ.
(Dallas, Texas)
and
SNR DENTON, LLP
BY: MARC S. FRIEDMAN, ESQ.
(New York, New York)
and
SNR DENTON, LLP
BY: ANDREW M. GRODIN, ESQ.
(Short Hills, New Jersey)
Counsel for Personalized User Model, LLP

12
13
14
15
16
17
18
19
20
21
22
23
24
25

POTTER ANDERSON & CORROON, LLP
BY: RICHARD L. HORWITZ, ESQ.
and
QUINN EMANUEL URQUHART OLIVER & HEDGES, LLP
BY: CHARLES K. VERHOEVEN, ESQ.,
DAVID A. PERLSON, ESQ., and
ANTONIO R. SISTOS, ESQ.
(San Francisco, California)
and
QUINN EMANUEL URQUHART OLIVER & HEDGES, LLP
BY: JOSHUA LEE SOHN, ESQ.
(Washington, District of Columbia)
and
QUINN EMANUEL URQUHART OLIVER & HEDGES, LLP
BY: ANDREA PALLIOS ROBERTS, ESQ.
(Redwood Shores, California)
Counsel for Google, Inc.

620
1

- oOo -

2

P R O C E E D I N G S

3
4

(REPORTER'S NOTE:

The following jury trial

proceedings were held in open court, beginning at 8:33 a.m.)

5

THE COURT:

6

(The attorneys respond, "Good morning, Your

7

Honor.")

8
9

Good morning.

THE COURT:

Are there any more issues plaintiff

wants to raise this morning?

10

MR. NELSON:

11

THE COURT:

12

I don't believe so, Your Honor.
Okay.

Any issues defendants want to

raise?

13

MR. PERLSON:

14

THE COURT:

15

MR. PERLSON:

Yes, Your Honor.

Good morning.

Good morning.
A few issues on, there are some

16

issues with some of the issues with the Konig exhibits.

I

17

don't think we're going to get to the, I guess the cross of

18

the cross for Konig.

19

and focus on Pazzani.

I think we can put that off for now

20

THE COURT:

Okay.

21

MR. PERLSON:

So there are a few different

22

objections for Pazzani.

There are two correspondence,

23

attorney correspondence between our firm and SRI Denton

24

that plaintiff seeks to introduce into evidence through Dr.

25

Pazzani basically about discovery disputes.

It's PTX-226

621
1
2

and 379.

And I think these are implicated by slide 109.
I mean it's very prejudicial to, and confusing

3

for the jury to get these things.

4

there is back and forth about discovery.

5

talking about how David Perlson isn't traveling today and

6

can't meet and confer, and then, you know, back and forth

7

about how your accusations are inappropriate.

8

reason to put this stuff in because they have the evidence

9

that they supposedly wanted for it elsewhere.

10

I mean, for example,
You know, it's

There is no

For example, PTX-266, they want to put in the

11

portion of the letter where Andrea Roberts, our counsel,

12

says if DoubleClick system does not provide adequate

13

advertisement, then YouTube is treated as any other

14

publisher in the Adsense for Content system.

15

to use her statement to prove this, but they played Nemeth's

16

testimony yesterday and said the same thing.

17

even need it.

18

And they want

So they don't

Then there is another one that they want to

19

put in that says, so YouTube doesn't do any special cookie

20

related targeting, but we provided the exact same thing in

21

our second set of supplemental interrogatory responses to

22

plaintiff's fourth set of interrogatories.

23

you know, irrelevant I guess, you know, really hearsay

24

statements from counsel, I mean I think the jury is going

25

to be very confused and it's totally unnecessary and

So having this,

622
1

irrelevant.

2
3

THE COURT:

We'll hear plaintiff's

response.

4
5

Okay.

MR. NELSON:
third page.

6

May I have slide 109, please?

The

Thank you.
So, Your Honor, this is slide 109.

This is the

7

first of the two letters we're talking about.

And we're not

8

seeking to introduce these to talk about discovery disputes

9

or anything like that.

We're seeking to introduce these

10

because these letters were provided to us in lieu of taking

11

additional discovery to try to remedy discovery disputes.

12

We relied on these letters, Dr. Pazzani relied on these

13

letters.

14

We offered last night, if they're willing to

15

stipulate that YouTube is treated as any other publisher in

16

the Adsense For Content system, because that is how we're

17

basically going to try to prove YouTube infringing because

18

it does the same thing as any other publisher in the Content

19

Ad system, if they're willing to stipulate to that, we will

20

withdraw this letter.

21

But if they're not, they introduced some

22

testimony counterdesignated from Nemeth that indicated maybe

23

they're still trying to differentiate the two systems.

24

this statement was provided in lieu of a 30(b)(6) deposition

25

to us, and we relied on that statement.

And

623
1

One, it's not hearsay.

It's an admission of

2

Google.

Two, it's highly probative.

3

prejudicial and we're not seeking to use the other stuff.

4

mean this isn't about discovery disputes.

5

THE COURT:

Three, it's not
I

But you want to offer into evidence

6

the underlying letter which I guess is PTX-226.

7

ahead and redact that and limit it just to the portion you

8

have on the slide here?

9

MR. NELSON:

10

We absolutely could, Your

Honor.

11
12

Sure.

Can we go

THE COURT:

And that's one of the two issues I

think raised.

13

MR. NELSON:

Yes, that is one of the letters.

14

May I have the other letter, please?

15

the second deck, please.

It is slide 183.

16

So while the slide is come up, Your Honor.

17

It's

So this one, let me just read a portion the

18

letter.

19

Christian Samay of our firm relating to a potential 30(b)(6)

20

deposition.

21

letter briefs, we think this should moot much, if not all,

22

of what PUM is seeking through interrogatory responses or

23

30(b)(6) depositions.

24
25

This is a letter from Google's attorneys to

And the letter says:

Given the timing of the

Then it goes on for three pages to identify
specific subroutines and code that link up to the Google

624
1

systems, so basically to put the code files together with

2

the functionality of the accused systems by name.

3

And, again, we absolutely relied on this letter

4

as a basis for which code does which thing for which system.

5

And so, you know, to the extent this letter has anything

6

about back and forth and underlying discovery stuff, we're

7

happen to redact that.

8

prejudice and some kind of discovery fight.

9
10

We're not looking for, you know,

THE COURT:

Are these letters referenced in Dr.

Pazzani's expert report?

11

MR. NELSON:

12

THE COURT:

13

MR. NELSON:

14

Yes, they are.
Do you have anything else?
No.

think they should come in.

15

THE COURT:

16

MR. PERLSON:

17
18

They're not hearsay and we

Mr. Perlson, do you want to respond?
Yes.

Can you put 109 back up,

please?
So I don't think -- what we're saying here,

19

first of all, if you can see on here, that there is

20

discovery squabble.

21

without merit.

22

issues are not ripe for the Court.

23

THE COURT:

24
25

The request for another 30(b)(6) is

And then on the bottom, they say, you know,

So if they redact all of that, do

you still have an objection?
MR. PERLSON:

I do.

And he is mischaracterizing

625
1

what it said.

2

very same testimony that they played yesterday.

3

We're pointing to, we're just repeating the

THE COURT:

4

on that?

5

All right.

But is there a dispute

YouTube is treated --

6

Are you disputing the substance of this point?

MR. PERLSON:

7

yesterday.

8

No, but he testified to it

correspondence.

9
10

There is no reason to put in attorney

THE COURT:

Are you willing to stipulate to it?

MR. PERLSON:

That there are two methods -- I

11

mean I will go back and see if there is some clarification

12

on that but I suppose if you want us to stipulate on

13

something we can do that.

14

THE COURT:

I don't want both --

15

MR. PERLSON:

I don't think we have to stipulate

16

to anything just because they tried to get in evidence

17

through a letter between us.

18

THE COURT:

You don't have -- I'm not going to

19

force you to stipulate to it, but I'm going to let them

20

either put in evidence that they relied on in lieu of

21

discovery, which it sounds like this is.

22

to eliminate hopefully any undue prejudice to you.

23

a disputed point, then they're entitled to put in the

24

evidence their expert relied on.

25

point, then it's not a disputed point, and my ruling would

We would redact it
If it's

If it's not a disputed

626
1

be different.

2

MR. PERLSON:

Just because our expert relied on

3

it doesn't mean it's admissible.

4

we're saying they're not required to, they shouldn't get

5

another 30(b)(6) deponent because we provided a deponent who

6

already testified to something and they played that

7

testimony.

8
9

What is in there is our --

They don't need this at all.
THE COURT:

All right.

Well, my ruling is I

guess you are not -- you can have time to think about

10

whether you want to stipulate, but if you don't stipulate,

11

then I will allow them to put on a redacted version of this

12

slide that takes out any confusion or risk of confusion

13

about there is a discovery dispute and what lawyers are

14

arguing about.

15

need to be redacted consistent with that.

16

Similarly, the underlying document would

The expert has apparently indicated in his

17

report he relied on this letter.

18

relied on it in lieu of taking additional discovery and for

19

those reasons, I would allow them to present the slide and

20

the underlying document.

21

MR. PERLSON:

22

And then the other one, 379, is this letter,

23
24
25

PTX-379.

Okay.

Plaintiff indicate they

We'll confer.

Could you -- I don't know if you can put that up.
I mean they say they relied on this letter, but

there was no rebuttal to my other, my main point which is

627
1

that we put this information in defendant's second

2

supplemental set of interrogatories and that same

3

information is in there.

4

And then we supplemented the interrogatory.

5

they should be pointing to.

6

THE COURT:

So we said it in a letter, sure.
That is what

They don't need this.

So if they redact all the back and

7

forth in discovery and, you know, you are unavailable, et

8

cetera, then what is the prejudice to them using it in a

9

letter form as well as in lieu of an interrogatory?

10

MR. PERLSON:

I don't think it's appropriate to

11

be putting before the jury a letter between counsel at all.

12

I mean it should be an interrogatory response.

13

type of thing that they could do for Your Honor.

14

it will be confusing for the jury to see the Quinn Emanuel

15

letterhead on there and have back and forth who are these

16

people, what is this thing behind them, when they're

17

getting, all it is is this information.

18

appropriate piece of evidence that should be before a jury.

19

THE COURT:

20

MR. PERLSON:

21

THE COURT:

That is the
I think

That is not an

Is there anything else?
No, Your Honor.

Right.

22

not just the interrogatory?

23

Not on that.

Understood.

Mr. Nelson, why

et cetera?

24
25

MR. NELSON:

Why do we need the letterhead,

Because the interrogatory doesn't

contain every single piece of information that this letter

628
1

does.

And I can't go back and point code by code file, but

2

this letter was given to us in response to a discovery

3

dispute, answering a lot of the questions that we relied on

4

it to.

5

THE COURT:

All right.

I'll accept your

6

representation.

You will need to redact it to take out

7

anything that night be characterizing counsel's interactions

8

and just limit it to the substance that you are going to

9

rely on.

But the fact that the jury will learn lawyers do

10

communicate and have a letterhead I don't think will be

11

unduly confusing or prejudicial under these circumstances.

12

MR. NELSON:

13

THE COURT:

Thank you, Your Honor.
So the objections are overruled.

14

you work out some stipulation, fine.

15

If

rulings.

16
17

MR. PERLSON:

THE COURT:

19

MR. PERLSON:

21

Again, obviously we need to see

the redacted copy.

18

20

Otherwise, you have my

Of course.

Of course.

Another one is PTX-381, slide 150.

Would you pull that up, please.
So this is a deposition of Google -- of a Google

22

witness, Michael Jahr in another case.

And I guess they

23

just want to read it to the jury, but it was on their

24

exhibit list.

25

and we didn't have a problem with them using the deposition

The actual deposition was on the exhibit list

629
1

just like they would any other deposition, but just because

2

the deposition is in another case doesn't mean that they

3

can just throw it up there.

4

deposition, then they should have designated it and, you

5

know, played it and we can counter and that sort of thing.

6

This is just taking an answer that they plucked out that

7

they want to show to the jury.

8

appropriate and it's not supposed to be how you use a

9

deposition.

If they want to use it like a

I don't think that is

10

THE COURT:

All right.

Thank you.

11

MR. NELSON:

12

So we tried to work this out last night, and I

Thank you, Your Honor.

13

guess we still haven't, but the only portion -- this is the

14

deposition from a previous case, not this case, and the only

15

portion of it that we want to use is what is on the slide,

16

not the whole deposition or anything like that.

17

And, you know, because it was from another case,

18

it basically just got left off the designations inadvertently;

19

and, you know, there is certainly a lot of law that says there

20

is no reason why this isn't admissible, you know, other than

21

it didn't come in in the form that it would have come in a

22

couple days ago.

23

We're happy if they want to counterdesignate

24

something, to have that come in as well on cross or just be

25

read in.

We just ask that this come in, in this limited way

630
1

as stated on the slide.

2

THE COURT:

All right.

Mr. Perlson, would you

3

like the opportunity to counterdesignate if I'm going to

4

allow them to do this?

5

MR. PERLSON:

No, Your Honor.

I mean I don't

6

think we should have to.

7

of this deposition is, frankly.

8

ruled they can't do this.

9

That they wanted to throw a bunch of deposition testimony

I don't know what the full scope
But Your Honor already

This is what we raised before.

10

up without designating it; and this is the exact -- it's

11

exactly what they've done.

12

another case doesn't excuse it from your prior ruling.

13

would essentially undo your prior ruling and there is, you

14

ruled they had a time to designate the testimony.

15

didn't and it shouldn't be admitted.

16

THE COURT:

And this, just because it's in
This

They

Is there other -- I don't know

17

what's still coming from you, but is any of it, from your

18

review of the slide deck, is there other deposition

19

testimony that would fall into this category where they

20

neglected to timely designate it?

21
22

MR. PERLSON:

I don't think so, as far as I

know.

23

THE COURT:

All right.

Well, I will sustain

24

this objection given the failure to comply with the guiding

25

order here.

It is only a couple of lines, but I'm not

631
1

persuaded that it's neglect that is, under the circumstances

2

it's usable in the form of prejudice to Google by allowing

3

it to be, so I will sustain that objection.

4
5

MR. PERLSON:

One final objection.

If you could

pull up PTX-1268, the replacement, I guess.

6

So here's -- we got a replacement set.

I guess

7

there was an earlier one before last night and if you could

8

just kind of page through it.

9

of stuff.

It's basically a hodgepodge

It's like search results and portions of Google

10

documents and stuff, you know.

11

that this goes back to the jury, and I don't think that's

12

appropriate.

13

And apparently the idea is

It's not a single exhibit.
If they wanted to mark some of this stuff and

14

introduce it as exhibits, that's fine.

15

other places, or if they just want to use it as a

16

demonstrative, that's okay.

17

Some of it is in

But it's inappropriate to just dump a bunch of

18

stuff in.

19

report.

20

but it shouldn't be an exhibit.

21

It's basically all the figures in Dr. Pazzani's

If they want to use demonstratives, that's fine,

THE COURT:

And is this an exhibit that was

22

disclosed to you consistent with the timing and all the

23

exchanges in connection with the pretrial order?

24
25

MR. PERLSON:

Well, there was a prior version of

it that was disclosed and it was objected to.

632
1

THE COURT:

And it was different than --

2

MR. PERLSON:

It was different than this.

I

3

don't know exactly how different.

We just got this last

4

night.

5

it came up before, but, you know, it seems to me there

6

should be a way to work this out because, again, they should

7

be able to use these as demonstratives, and if there are a

8

couple parts of it that they really think need to go back to

9

the jury -- like I think Mr. Nelson had mentioned that the

And, you know, out of 90 exhibits, I don't know if

10

first two are like search results from something that Dr.

11

Pazzani did, totally fine with that.

12

that we should just throw all of this in.

13

THE COURT:

14

MR. NELSON:

All right.

We just don't think

Mr. Nelson?

And we may be able to work this

15

out.

16

supposed to be all the figures in Dr. Pazzani's report and

17

somehow some SRI-related stuff got attached to it.

18

not intended.

19

I mean, what happened was the original 1268 was

That was

That was disclosed in a timely manner.

We then were attempting to fix that, and in

20

fixing that, apparently some of the figures got left out.

21

That was the confusion at the start of his testimony

22

yesterday, just the picture of what search results are, you

23

know, Search Ads.

24

could just put them up, put some of the figures up as a

25

demonstrative now as we go through this so we can move

And so if on their representation, if we

633
1

through Dr. Pazzani's examination efficiently, we can then

2

probably work out the portion that gets marked as exhibits

3

after the fact.

4
5
6
7
8
9

THE COURT:

It sounds like that's a workable

arrangement to at least get through the testimony.
MR. PERLSON:

Yes.

Yes.

We just want to make

sure that demonstratives don't go back to the jury.
THE COURT:

Right.

So the demonstratives don't

go back to the jury just because they are demonstratives,

10

but you all need to talk about which subset of what has been

11

marked as PTX-1268 the plaintiffs want to be admitted into

12

evidence and go back to the jury.

13

about that, and if we have disputes, we'll deal with them

14

some time before we submit the case to the jury.

15

MR. NELSON:

16

THE COURT:

17

MR. NELSON:

18

THE COURT:

19

MR. PERLSON:

20

THE COURT:

21

MR. VERHOEVEN:

And you will have to talk

Yes.
Understood?
Yes, Your Honor.
Understood?
Understood.

Any other issues from defendant?
No, Your Honor.

I just want to

22

put on the record, Mr. Pazzani's deposition, several times

23

when he was asked questions, he said basically, I wanted

24

this, but Google wouldn't give me the documents.

25

and conferred because we thought that was inappropriate for

And we met

634
1

him to say such a thing in front of the jury and we reached

2

agreement that he is not going to say that.

3

to put it on the record just in case something happens.

4

THE COURT:

5

MR. VERHOEVEN:

6

But I just want

reached agreement on that.

7

THE COURT:

8

MR. NELSON:

9

THE COURT:

All right.
My understanding is we have

Is that correct?
My understanding is the same.
And I think Dr. Pazzani is in the

10

courtroom, but, Mr. Nelson, have you advised him of this

11

agreement and/or will you?

12
13

MR. NELSON:
he is now so advised.

14
15

I think I have, but if I haven't,

THE COURT:

All right.

Anything else from

defendant?

16

MR. VERHOEVEN:

17

THE COURT:

No, Your Honor.

What are the proposals on submitting

18

jury instructions and verdict sheet?

19

MS. JACOBS:

Your Honor, I need to hear back

20

from our, how long it's going to take to compile this into a

21

single document.

22

able to do that today, and we have not -- based on that, we

23

need to talk about when we can meet and confer and get that

24

done.

25

We have not -- my assumption is we'll be

My hope would be by Friday.
THE COURT:

It's certainly going to have to be

635
1

by Friday that we get something.

2

Mr. Horwitz?

3

MR. HORWITZ:

That's my understanding.

Once

4

they're compiled into one document, then each side needs to

5

look at them.

6

we can resolve some differences from where we were before

7

the pretrial conference and the aim is to get them to you

8

during the day on Friday.

9

I think we need to spend some time to see if

THE COURT:

Okay.

Well, let's revisit it

10

tomorrow.

Let's say they will be due Friday.

11

tomorrow I will set a time on Friday, but I will hear from

12

you first as to what your proposal is as to what time on

13

Friday it should be.

Okay?

14

MS. JACOBS:

15

THE COURT:

16

MR. PERLSON:

17

Sometime

Thank you, Your Honor.
Yes?
One thing I'm happy to report in

agreement between the parties.

18

THE COURT:

We like agreements.

19

MR. PERLSON:

On, I mentioned the issue of JMOLs

20

yesterday and the parties agreed, and if it's acceptable to

21

Your Honor, we would propose that we do JMOLs orally, argue

22

them, and then the parties have the option to follow it up

23

with a written brief, if that's acceptable.

24
25

THE COURT:

Certainly, it's acceptable.

Do you

have an agreement as to how long you will give each other to

636
1

file the written brief?

2

MR. PERLSON:

We had suggested three days after.

3

We didn't, we didn't agree to limit ourselves either way, I

4

guess, but if you want us to limit the time after, that

5

would be my suggestion, but I'm sure we can work that out.

6

THE COURT:

All right.

Well, I think there

7

should be some sort of deadline to it, so see if you can

8

continue the agreement trend and work out an agreement on

9

that and let me know at some point --

10

MR. PERLSON:

11

THE COURT:

12

MR. PERLSON:

13

-- what that agreement is.
Do you have any preference on how

much time after?

14

THE COURT:

15

MR. PERLSON:

16

THE COURT:

17

Okay.

I think the sooner, the better.
Okay.

But something in the neighborhood of

three days sounds fine, but I'm not holding you to that.

18

MR. PERLSON:

19

THE COURT:

20

MR. PERLSON:

21

THE COURT:

Okay.

You talk.
Fair enough.

All right.

22

check in with the jury?

23

MS. JACOBS:

24

MR. PERLSON:

Anything else before we

25

agreement.

No, Your Honor.
One more thing.

Another

637
1

THE COURT:

You're the bearer of good news.

2

MR. PERLSON:

On source code I had mentioned

3

that we wanted to, that there were some confidential

4

portions that we had concerns.

5

carefully and it is highly confidential stuff, but it's our

6

understanding it's only going to be shown for a very short

7

period of time and that, you know, we could deal with it

8

later in terms of the transcript and work with plaintiffs in

9

terms of redacting it.

We looked through it very

But we understand it's a burden on

10

the Court to open and close it, and so we decided that we

11

won't -- we don't need to close the courtroom.

12
13

THE COURT:

And does that go for today or for

the remainder of trial, as far as you know?

14

MR. PERLSON:

Well, I don't know exactly what

15

will come in later, but it goes for Dr. Pazzani's

16

presentation.

17

THE COURT:

Okay.

So we will not anticipate

18

being asked to close the courtroom during Dr. Pazzani's

19

testimony.

20

MR. PERLSON:

21

THE COURT:

22

Correct.

All right.

Is that correct from

your point of view as well?

23

MR. NELSON:

That is correct.

24

THE COURT:

All right.

25

All right.

We'll take a short break and then

Anything else?

638

Pazzani - direct
1

we'll get the jury ready.

2

(Short recess taken.)

3

THE COURT:

4

(The jury entered the courtroom.)

5

THE COURT:

We will bring the jury in.

Good morning, ladies and gentlemen

6

of the jury.

7

think we'll have Dr. Pazzani return to the stand.

8

Good morning and welcome back to you.

9

Welcome back.

We are prepared to begin, so I

I remind

you, you remain under oath.

10

Mr. Nelson, you may proceed.

11

... MICHAEL PAZZANI, having been previously sworn

12

as a witness, was examined and testified further as follows ..

13

DIRECT EXAMINATION (Continued)

14

BY MR. NELSON:

15

Q.

Good morning, Dr. Pazzani.

16

A.

Good morning.

17

Q.

Let's go back briefly to yesterday to sort of

18

reorient people to where we left off.

19

A.

20

Okay.
MR. NELSON:

Could I get slide three, please?

21

BY MR. NELSON:

22

Q.

23

is this slide?

24

A.

25

can type some terms in the search box like car repair, click

I want to talk just about the accused products.

This slide shows the Google Search system.

What

A user

639

Pazzani - direct
1

on Google Search and get search results.

2

MR. NELSON:

Can I have slide 4, please in.

3

BY MR. NELSON:

4

Q.

Can you identify the search results?

5

A.

Yes.

6

number of search results here.

7

green in the lower center.

8

car advisor web results about do-it-yourself auto repair.

9

Q.

This is a search results page.

There are a

They're in the box that's

It contains things like expert

And what about the portion labeled Search Ads?

What

10

is that?

11

A.

12

search results page.

13

users click on the ads and Google conducts an auction to

14

decide which ads to display there based on part of

15

probability the user will click on the ad and on how much

16

the advertiser is willing to pay for it.

17

Q.

18

case?

19

A.

The Search Ads are ads that are displayed on the
Advertisers offer to pay Google when

And are these two of the accused products in this

Yes, they are.

20

MR. NELSON:

And let me have slide 5, please.

21

BY MR. NELSON:

22

Q.

And can you explain what this slide is?

23

A.

Yes.

24

advertisers can also display ads on content pages, things

25

like CNN or the Los Angeles Times or blogs.

This is an example of Content Ads.

So

640

Pazzani - direct
1

So this is actually my blog.

I'm a bird watcher

2

and I've entered into an agreement with Google to display

3

ads on the blog, and the ads are related to birds.

4

similar auction is conducted to decide which ads to display

5

there.

6

Q.

7

yesterday.

8

A.

9

video or click on the ad about the Fiat that's in the bottom

Let me have slides 5 and 6, please.

A

We had this up

This is YouTube?

Yes.

This is a YouTube video.

Users could watch the

10

of the screen.

11

Q.

12

those the other two accused products in this case?

13

A.

That's correct.

14

Q.

So you gave a summary of your opinions yesterday and

15

I won't repeat that because I want to move on to what makes

16

you qualified to give an opinion in this case?

17

me your educational background?

18

A.

19

computer science from the University of Connecticut.

20

Master's was in 1981 and I specialized in artificial

21

intelligence.

22

Are Content Ads and the YouTube Content Ads, are

Yes.

Can you tell

I received a Bachelor's and Master's degree in
The

I received a -- then I went to UCLA, University

23

of California, Los Angeles, where I received my Ph.D. in

24

1998 in computer science specializing in machine learning.

25

From there I became a Professor of Information

641

Pazzani - direct
1

and Computer Science at the University of California,

2

Irvine, where I was on the faculty there for about 18 years

3

or so.

4

I went to the National Science Foundation in

5

Washington, D.C., where I oversaw the government's programs

6

in information intelligence systems, funding research in

7

areas like machine learning, databases, search engines, and

8

personalization.

9

From there I became the Vice President of

10

Research at Rutgers, the State University of New Jersey,

11

just up the road a bit.

12

And in 2012 I went back to California, at the

13

University of California, Riverside, where I'm now the Vice

14

Chancellor of Research and Economic Development and also a

15

Professor of Computer Science and Engineering.

16

Q.

Thank you, Dr. Pazzani.

17

Let me ask you a couple other questions.

So you

18

mentioned you had involvement with the National Science

19

Foundation.

Can you explain that a little bit further?

20

A.

At the National Science Foundation, faculty

21

from across the country submit proposals and ideas that they

22

would like to do, and the National Science Foundation finds

23

about 15 percent of them to be very meritorious.

24

provides funding so that the faculty can conduct that

25

research to advance the state of the art, areas like machine

Yes.

And that

642

Pazzani - direct
1

learning or personalization, search engines, databases,

2

speech recognition, many of the topics involved in this

3

case.

4

Q.

5

things like that?

6

A.

7

Artificial Intelligence.

8

are elected by our peers due to their accomplishments; and I

9

think I was elected in about 2004.

And did you have other awards such as being a fellow,

Yes.

10

So I'm a fellow of the American Association of
Approximately, 10 people a year

I'm also on the editorial on the Machine

11

Learning Journal and a couple other journals.

12

Q.

Have you taught courses in machine learning?

13

A.

Yes, I have taught numerous courses in artificial

14

intelligence, machine learning, personalization.

15

Q.

Have you published in this area as well?

16

A.

Yes, I have a number of publications in that area.

17

MR. NELSON:

Let me have slide 10, please.

18

BY MR. NELSON:

19

Q.

20

to prepare yourself to give an opinion in this case?

21

A.

22

that were stored on a secure website.

23

look through those documents.

24

printed out then because they contained words related to

25

this case.

So can you explain to the jury what you did in order

Yes.

So Google made available thousands of documents
And I was able to

Approximately 250 were

And I read those in detail.

643

Pazzani - direct
I attended the depositions of many Google

1
2

witnesses.

Those were the videos that you saw yesterday.

3

And I also read the depositions of those that I could not

4

attend.
PUM's lawyers asked Google its questions,

5
6

interrogatories that you heard about.

7

responses to those interrogatories.

8
9

And I read Google's

I spent about 30 days in front of computers in
San Francisco, New York and Los Angeles in Google's lawyers

10

offices studying the source code for Google trying to

11

understand how it operates, and you will see just a little

12

bit of that today.

13

And then I also read numerous public documents

14

regarding Google systems, their help files or their box that

15

explains their features.

16

Q.

17

systems.

18

A.

Yes, I did.

19

Q.

And can you explain those experiments?

20

A.

Yes.

21

of users.

22

lover and recreate an account called Brianne animals; and

23

she searched for animals, cats, dogs, birds and click on web

24

sites related to those animals or click on ads related to

25

animals.

And did you conduct any experiments using the Google

I asked an assistant to simulate various types

So in one case, I asked her to simulate an animal

644

Pazzani - direct
1

In another case, I asked Brianne to simulate a

2

user who was a car enthusiast, and so she would search for

3

things like Mustang or Honda Civic and click on the search

4

results related to Mustangs and Honda civics.
And then Google provided the profiles, the

5
6

information that Brianne had done over the course of a year,

7

and we got together about one once a month.

8

review the history of what she did and look at the search

9

results of the two accounts and compare them.
MR. NELSON:

10

I was able to

May I have slide 11, please.

11

BY MR. NELSON:

12

Q.

13

the Brianne car accounts that are on the slide?

14

A.

15

discussing.

16

Q.

And can you turn to your binders, PTX-375, please?

17

A.

Is it the big one?

18

Q.

The big one, yes.

19

A.

(Witness complies.)

20

Q.

Can you tell me --

21

A.

That was 373.

22

Q.

375 may be the really big one.

23

A.

Yes, 375 is the really big one.

24

Let's move over here.

25

And are those two accounts, the Brianne animals and

Yes.

Those are the two accounts that we'll be

Sorry.

Yes, this is PTX-375.

That's part 2.

645

Pazzani - direct
1

Q.

And can you just identify PTX-375 for the jury, what

2

it is?

3

A.

4

So this contains, all these two binders contain all of the

5

searches that she did over the course of about eight or nine

6

months in 2011 and every search result, everything she

7

clicked on, every ad she saw, every ad she clicked on as

8

well as the profiles that Google took of her interests.

9

Q.

PTX-375 is the profile of the Brianne cars account.

And that is information that came from

10

correct?

11

A.

12

actually checked this morning.

Yes,

.

And I

13
14
15
16
MR. NELSON:

17
18

I'd like to offer PTX-375 into

evidence.

19

MR. VERHOEVEN:

No objection.

20

THE COURT:

21

(PTX-375 admitted into evidence.)

It's admitted.

22

BY MR. NELSON:

23

Q.

Please identify PTX-373.

24

A.

PTX-373 is the profile of the Brianne animal

25

accounts.

And, similarly, it contains the animal searches

Pazzani - direct

646

1

done using this account, simulating an animal lover as well

2

as the ads she clicked on, the ads that she saw in 2011.
MR. NELSON:

3
4

And I'd like to offer PTX-375 --

373 into evidence.

5

MR. VERHOEVEN:

6

THE COURT:

7

(PTX-373 admitted into evidence.)

8

Q.

It's admitted.

BY MR. NELSON:

9

No objection.

Can you look at PTX-33 and 34 and just together

10

identify what those exhibits are?

11

A.

So PTX-33 and 34 are the Google

Q.

So together, PTX-33, 34, 37 --

12
13
14
15
16
17
18

MR. NELSON:

I'd like to offer PTX-33 and 34

into evidence.

19

MR. VERHOEVEN:

20

THE COURT:

21

(PTX-33, PTX-34 admitted into evidence.)

22

Q.

24

information that was

Those are admitted.

BY MR. NELSON:

23

No objection.

25

So together, PTX-33, 34, 373 and 375, all of that is

correct?

647

Pazzani - direct
1

A.

That is correct.

2

Q.

And so this is all the information that Google has

3
4

A.
MR. NELSON:

5

Let's turn to a little bit of

6

history and kind of go back in time for a little while.

7

I get the next slide, please.

8

BY MR. NELSON:

9

Q.

Can

And I want to talk but World Wide Web and the

10

Internet.

Did you prepare a demonstration or an

11

illustration to discuss early search engines?

12

A.

13

interpret was like before search engines.

Yes, I did.

First, let me just explain what the

So before there were search engines, to go to a

14
15

website, you had to type an URL.

16

WWW.NewJersey.com/CapeMay.html to see about my bird watching

17

at Cape May.

18

Something like

But it's really hard to remember those URLs so

19

search engines were made to make it easier for people to

20

find things.

21

MR. NELSON:

Can I have slide 13, please.

22

BY MR. NELSON:

23

Q.

And can you explain what is going on in slide 13?

24

A.

Yes.

25

overview of what a search engine is.

Slide 13 shows a search engine.

It's a brief

Essentially what a

Pazzani - direct

648

1

search engine does is first it has a program called a

2

crawler that goes out on the web and finds all of the

3

documents on the web.

4

document, finds the important words in those documents and

5

puts them in a database.

6

words with the address, the URL, so that when you type one

7

of those important words, it can find that document on the

8

web.

And it actually analyzes those

It associates those important

So that's one part of the search engine.

9
10

the part that is stored at the search engine company's

11

That's

headquarters.

12

The part that you use, that users use is the

13

query box, if you like.

14

engine, a word like jaguar.

15

the database, finds all the counts that contain the word

16

"jaguar" and it returns some of those documents as search

17

results to the user and on a search result page that

18

contains links to the documents on the website like the New

19

Jersey birds website.

20

MR. NELSON:

You type a query into a search
The search engine looks through

Let me have slide 14, please.

21

BY MR. NELSON:

22

Q.

23

depict?

24

A.

25

is getting search results about jaguar, but this particular

Can you just explain what this slide is intended to

Well, this slide is intended to depict that the user

649

Pazzani - direct
1

user, a car enthusiast is much more interested in the car

2

pages about jaguar, but they're hidden among the sports

3

pages, the Jacksonville Jaguars or the animal pages or even

4

the Jaguar guitar pages, so he can't exactly find what he is

5

looking for.

6

MR. NELSON:

Let me have the next slide, please.

7

BY MR. NELSON:

8

Q.

And what is this slide intended to illustrate?

9

A.

This slide is illustrating the general principle of

10

information overload.

11

contain the word "jaguar" and how do you find the few that

12

are the most important, the most interesting?

13

the problem that Konig and others tried to solve.

14

There might be 22,000 documents that

MR. NELSON:

And that's

And let me go to slide 18.

15

BY MR. NELSON:

16

Q.

And did that problem have a name?

17

A.

Basically, the name is information overload.

18

just hard to find what you are looking for.

19

trying to find a needle in a haystack.

20

find the jaguar car dealer, you might find jaguars at the

21

local zoo instead.

22

Q.

23

that problem become more serious?

24

A.

25

available, it's harder to find the document that you are

It's

It's like

If you are trying to

As the Internet grew and the World Wide Web grew, did

Sure.

So as there were more and more documents

Pazzani - direct
1

looking for.

2

Q.

3

turned in your binder to PTX-6, please.

4

A.

(Witness complies.)

5

Q.

And can you just identify what that is?

6

A.

Yes.

7

entitled Souped-Up Search Engines.

8

Q.

9

650

is that right?

10

A.

11
12

Let me take you now to 2000 and let me have you

Yes.

This is an article from Nature Magazine

And the founder of Google is quoted in that article;

That's correct.
MR. NELSON:

Your Honor, I'd like to offer PTX-6

into evidence.

13

MR. VERHOEVEN:

No objection, Your Honor.

14

THE COURT:

15

(PTX-6 admitted into evidence.)

16

MR. NELSON:

It's admitted.

Can I get slide 19, please.

17

BY MR. NELSON:

18

Q.

19

which is May 11, 2000.

20

future of search at this time?

21

A.

22

five years, the search engine as we know it will no longer

23

exist, or be marginal.

24

programs that search by using their experience of the needs

25

and interests of their users.

So let's fast forward now to the date of this article
What did Mr. Brin say about the

The article said:

Google's Brin predicts that in

In its place will come intelligent

651

Pazzani - direct
MR. NELSON:

1

And let me turn now to the patents

2

in question.

Let me have slide 21.

3

the patent.

4

BY MR. NELSON:

5

Q.

And so can you identify what is on slide 21 or 22?

6

A.

Slide 22 is the '040 and the '276 patents, entitled

7

Automatic, Personalized Online Information and Product

8

Services.

9

And it was based on a provisional patent filed in December

Actually, 22.

Go to

Thank you.

The inventors are Konig, Twersky and Berthold.

10

1999.

11

Q.

And these are the patents in this case; right?

12

A.

That's correct.

13

Q.

And they were filed in December 28th, 1999; right?

14

A.

The provisional was.

15

Q.

Yes.

16

Mr. Brin made his statements in that Nature article; isn't

17

that right?

18

A.

Yes, it is.

19

Q.

So one of the things that Dr. Konig talked about was

20

learning machines and machine learning yesterday, and you

21

talked about it a little while ago in going through your

22

background.

23

what a learning machine actually is?

24

A.

25

that.

Yes.

And that is about five and-a-half months before

Did you prepare an illustration to help explain

We worked on a pretty simple illustration of

652

Pazzani - direct
MR. NELSON:

1

Let's begin with slide 24.

2

BY MR. NELSON:

3

Q.

And what is on slide 24?

4

A.

Well, on slide 24 is an apple.

5

trying to teach a child what an apple is or maybe teach a

6

computer what an apple is.

7

trying to do what human learning does but with a computer

8

instead of a person.
MR. NELSON:

9

But imagine you are

And really machine learning is

And let's have slide 25.

10

BY MR. NELSON:

11

Q.

Can you explain the graphic on slide 25?

12

A.

Yes.

13

On the left you will see a box with a funnel and a monitor

14

and we are going to put examples in that.

15

things like this is an apple, this is not an apple.

16

we're going to ask the machine to learn what an apple is.

17

Once it has learned, we're going to ask it to

So this graphically depicts a learning machine.

Examples of
And

18

apply its knowledge of what an apple is.

We're going to

19

show it things and ask it what an apple is.

20

what you might do with a child.

21

an orange.

22

will learn how to tell apples from oranges and pears, et

23

cetera.

24

Q.

25

a monitor.

This is sort of

This is an apple, this is

This is a green apple.

And over time, the child

And you have got this written sort of as a box with
What is in reality learning machines are

653

Pazzani - direct
1

mathematical functions and models?

2

A.

3

implemented by computer programs.

4

Q.

5

parts represent?

6

A.

7

part that you give examples with their answers to, and the

8

right part is the applying your knowledge part.

9

questions:

10

Yes.

They're mathematical functions, models often

And you have got two parts there.

What do the two

Well, the left part is the training part.

Is it an apple?

It's the

You ask it

And it will tell you whether

it's an apple or not.

11

MR. NELSON:

Let's go to the next slide now.

12

BY MR. NELSON:

13

Q.

Can you explain what is going on?

14

A.

Yes.

15

learning machine.

16

indeed an apple.

17

Q.

And go ahead and start the animation.

18

A.

And now the apple has gone into the learning machine

19

and it has learned something about the properties of apple.

20

Perhaps their shape, their color, they have stems, things of

21

that sort, and you can see it's depicted by an approximate

22

sketch of an apple.

23

Q.

And why is it kind of fuzzy at this point?

24

A.

Well, it doesn't fully know what an apple is yet.

25

has only seen one example.

So we're putting a labeled example into the
We have it labeled as an apple, and it is

It

654

Pazzani - direct
MR. NELSON:

1

Let's go to the next slide, please.

2

BY THE WITNESS:

3

A.

4

different shaped, maybe a little brighter red.

Okay.

This is another example of an apple slightly

MR. NELSON:

5

Can you please run this.

6

BY THE WITNESS:

7

A.

8

builds a slightly better model of what an apple is.

9

Q.

Now, what do we have?

10

A.

It looks like a lime, and we're telling it this is

11

not an apple.

12

with a lime and first think it's any small fruit and it gets

13

put into the machine and told, no, that is not an apple.

And the apple goes into a learning machine and it

14

And, again, your child might confuse an apple

MR. NELSON:

15

BY THE WITNESS:

16

A.

17

apple is.

You can go ahead and run that.

18

And it learns a little bit more precisely what an
Perhaps the color is more important.
MR. NELSON:

Let's go to the next slide, please.

19

BY THE WITNESS:

20

A.

21

asking it a question.

22

asking it the question is this an apple?

23

likelihood this is an apple?

24
25

So the next part shows the applying part.

It's learned something and we're

MR. NELSON:
BY THE WITNESS:

We're

Or what is the

And go ahead and run that one.

655

Pazzani - direct
1

A.

So the model of the apple has moved over and the

2

model looks at the properties of the apple, the shape, the

3

size, and it's asked a question:

4

answers, yes, indeed it is an apple and says there is a .92

5

likelihood or probability that this is an apple that is

6

above 50 percent.

7

pretty good one.

8

seen before.

9

Q.

And let's go to the next slide.

10

A.

Now we show it something else and we ask if this is

11

an apple and the properties are examined.

12

it's -- we ask the model whether it's an apple and it

13

produces a prediction.

14

chance it's an apple, so it's less than 50 percent.

15

yes -- so, no, this is not an apple.

16

Q.

And let's go to the next slide, please.

17

A.

Now we see something else and ask, is this an apple?

18

And the properties -- well, the model looks at the

19

properties, asked if it's an apple, and it says there's a

20

55-percent chance it's an apple.

21

aren't perfect.

22

this is an apple.

23

little more.

24

Q.

So let's go to the next slide, please.

25

A.

We put other examples in.

Is this an apple?

So, yes, this is an apple.

And it

Looks like a

Not exactly identified to what we have

It's generalized to learn something about it.

We ask whether

It says there's only a ten percent
So,

Our learning machines

There's a better than 50-percent chance
It's a tomato.

So we need to teach it a

We tell it a pineapple is

656

Pazzani - direct
1

not an apple, a green apple.

Grapes are not apples.

2

Slightly different apple.

3

don't even know what that is.

4

Q.

And so now what happened to the picture of the apple?

5

A.

It has become more precise.

6

has improved over time as we've given it more and more

7

examples.

8

Q.

9

that intended to represent?

Strawberries are not apples.

I

The model of the apple

It has learned.

And what is the, the apple on the screen, what is

10

A.

It's just intended to represent the learning

11

machine's model of an apple.

12

Q.

And what it knows about the apple?

13

A.

Yes.

14

Q.

And let's go to the next portion, the next

15

examination.

16

A.

17

the properties of the tomato and sees, for instance, that

18

the color differs, the stem shape is a little different.

19

And now it does a better job and says there's only an

20

18-percent chance this is an apple, less than a 50-percent

21

chance.

22

idea behind the machine learning.

23

mathematics behind it, but it's just presenting examples and

24

learning them.

25

Q.

So now when we give it the tomato, the model looks at

So, no, this is not an apple.

And that's the basic

There's lots of

And let's turn to -- now let's turn back to the

Pazzani - direct

657

1

patents and talk about the patents and the claims a little

2

bit.

3

A.

Sure.

4
5

MR. NELSON:

And to do that, may I set up a

couple of boards, Your Honor?

6

THE COURT:

7

And, Mr. Verhoeven, if you need to move around,

8

You may.

feel free to do so.

9

MR. VERHOEVEN:

10

Thank you, Your Honor.

(Mr. Nelson placed boards on the easels.)

11

BY MR. NELSON:

12

Q.

And, Dr. Pazzani, if I can ask if you step down --

13

A.

All right.

14

Q.

-- it might be easier to describe what's on here

15

standing up as opposed from over there.

16

(The witness left the witness stand and

17

approached the easels.)

18

BY MR. NELSON:

19

Q.

And can you just explain what is on the left?

20

A.

Yes.

21

the invention.

22

Q.

And it's Figure 2; is that correct?

23

A.

Yes.

24

Q.

And the color codes aren't in the original figure,

25

are they?

This is a figure in the patent that describes

658

Pazzani - direct
1

A.

No.

2

Q.

And then can you explain what is on the right?

3

A.

That's that same figure together with the claims of

4

the patent.

5

Q.

6

intend the color coding to represent the whole claim or just

7

the little estimate parameters, for example, on the bottom?

8

Go ahead and explain.

9

A.

And so when we're doing the color coding, do you

Yes, it may be parameters, represents the entire

10

claim, Element C.

You just can't fit the entire claim in

11

that little box, so it says, just summarizes it.

12

analysis, I will go through each of these words.

13

Q.

14

respective claim element?

15

A.

16

probabilities, corresponds to E, that statement.

17

Q.

But in my

And each of the colors then corresponds with the

Yes.

So 32 corresponds to B, and 38, estimating

Okay.

Thank you.

I will take these down.

18

(The witness resumed the witness stand.)

19

MR. NELSON:

And may I have slide 35, please?

20

35.

21

BY MR. NELSON:

22

Q.

23

let's start with analyzing the document.

24

you want to start there?

25

A.

And so let's start just explaining the patent, and

It's actually step 36.

Step 35.

Why did

659

Pazzani - direct
1

Q.

Oh.

2

A.

Analyzing the document is something that is not

3

dependent on the user.

4

particular claim depend on the user.

5

document analysis is the same for every user, so you might

6

want to do that once ahead of time instead of do that for

7

each user because the results of that analysis will be

8

exactly the same.

9

good idea to do it ahead of time so you have the results

10

All the other steps in the, in this
This is -- the

You can do it in any order, but it's a

when you need them.

11

MR. NELSON:

And let me have slide 36, please.

12

BY MR. NELSON:

13

Q.

14

what's on slide 36?

15

A.

16

within the patent.

17

analysis, finding a list of users previously interested in

18

the documents.

19

document in the past, for example.

20

And this is a portion of the patent.

Yes.

Can you explain

Slide 36 describes various types of analysis
So at the bottom is one type of

That's analyzing who has clicked on that

In the middle there's something I want to talk

21

about.

It's a topic classifier probability distribution.

22

That's just a fancy mathematical way of saying documents can

23

have multiple topics.

You can be partially about sports and

24

partially about cars.

For instance, a document that might

25

describe giving away a car at halftime of the basketball

660

Pazzani - direct
1

game to one of the audience member who shoots a hoop, that

2

would be partially about sports and partially about cars.

3

MR. NELSON:

And let me have slide 37, please.

4

BY MR. NELSON:

5

Q.

6

that slide, please?

7

A.

8

be displayed in hierarchies.

9

have a topic magazines, which is part of the topic

This is Figure 7 from the patent.

Yes.

Can you explain

So in Figure 37, it shows that the topics can
So you could, for instance,

10

publishing, which is part of the topic industry, which is

11

part of the topic business, the business of publishing

12

magazines, if you like.

And there's many of these topic

13

hierarchies on the web.

In the patents, they have looked at

14

the open directory project as one of the sources of

15

obtaining a hierarchy of topics.

16

MR. NELSON:

Let me have the next slide, please,

17

38.

18

BY MR. NELSON:

19

Q.

20

interactions with the data, or let's turn to -- did you

21

prepare an animation to illustrate these steps?

22

A.

23

user interactions mean.

24

MR. NELSON:

25

please.

And so now let's turn to monitoring the user's

Yes, I did.

So we can look at what does monitoring

And let me have the next slide,

661

Pazzani - direct
1

THE WITNESS:

So here's a user.

They type a

2

word like "dictionary" and then they go to the

3

Merriam-Webster website, and so monitoring the user is

4

watching the user as the user does those actions:

5

the query, receiving the results, and looking at the results

6

the user clicked on.

7

Q.

8

represent?

9

A.

Typing

Is that what the magnifying glass is intended to

Yes.

10

The magnifying glass is depicting things.
MR. NELSON:

And let me have slide 40, please.

11

BY MR. NELSON:

12

Q.

13

files.

14

set?

15

A.

16

just monitoring the user, it's recording what the user has

17

done.

So let's turn to the updating user-specific data
The next slide, please.

Yes.

And can you explain that

So updating the user data files, it's not

18

So this is Figure 14 from the patent, which

19

shows the user has gone to two websites, the Merriam-Webster

20

one we saw before as well as a herring.com website, and you

21

can see they're stored in a database.

22

contains the document ID, which would be in this case the

23

URL of the document.

24

you accessed it, how you got there, whether you were

25

interested.

That database

It also contains things like the time

It contains the context.

For instance, you

662

Pazzani - direct
1

typed the query dictionary to go to the Merriam-Webster

2

website.

3

So updating user-specific data files is

4

continuously doing things, continuously updating your data

5

about what the user has done.

6

Q.

7

patent; is that correct?

8

A.

9

like the URLs.

And is this, the top part, that's Figure 14 from the

Yes.

So the patent describes document IDs or things
And, again, they're storing the interaction

10

times, the access times and other items.

11

Q.

12

on the top part.

13

the top portion of the slide?

14

A.

15

auto member websites.

16

"Ford" or "Kelly Blue Book," and after doing that, they've

17

gone to those websites, and this is just updating the

18

database to include information about the user that has done

19

that.

And let's turn to slide 42, please.

Yes.

And let's focus

Can you explain what's going on now with

So this is a user who has visited a number of
They typed things like "used cars" or

20

The patent says that it's the user and his or

21

her associated, associated representation denoted with U.

22

In this case, that representation might be an identifier,

23

the user identifier that identifies that user.

24

Q.

25

were there for users?

And in 1999, in late 1999, what sorts of identifiers

Pazzani - direct

663

1

A.

Well, there are two common ways of identifying users.

2

One is having them create an account with a password.

3

Another is to put a cookie, a small piece of text on their

4

machine with an identifier.

5

Q.

6

tiny people inside the computers, did you?

7

A.

No.

8

Q.

Some sort of electronic representation?

9

A.

There's an electronic representation, yes.

10

Q.

And let me turn to slide 43.

11

A.

So slide 43 is intended to depict the fact that there

12

are multiple users of the system.

13

their own identifier, and there might be one large central

14

database in which all of this data is stored, but each user

15

has their own identifier.

16

search history by using her identifier and the man's search

17

history by using his identifier.

18

searching for animal-related topics and he has been

19

searching for car-related topics.

20

Q.

And so are the respective, let's take the one on the

21

left.

Is that then, is that information then associated

22

with the user ID, for example, on the left side of the

23

figure?

24

A.

25

database with a key, the user ID, and from that ID, you can

So in 1999, just like today, you didn't have little

Yes.

The people are represent by some number.

Each of them would have

And you can find the woman's

It looks like she has been

So one way to do that would be to have a

664

Pazzani - direct
1

get that user's interactions.

In this case, the man's

2

interactions with the car website.

3

Q.

4

document ID that you described earlier in the patent?

5

A.

6

representation of the document.

7

Q.

8

on the left side, www.Kelly, KBB for Kelly Blue Book,

9

www.AutoTrader, is that, is that a set of documents

And are documents also associated with the respective

Yes.

So it, in this case, it's storing the URL as a

And so together, that list of, that list of documents

10

associated with the user?

11

A.

Yes.

Related to the user ID.

12

MR. VERHOEVEN:

13

THE COURT:

Objection to leading.

Objection to leading.

I'm

14

sustaining it.

15

BY MR. NELSON:

16

Q.

Can you explain the column on the left?

17

A.

The column on the left are the documents the user has

18

visited and they're associated with the user by means of the

19

document ID and the user ID.

20

MR. NELSON:

Let me have slide 44, please.

21

BY MR. NELSON:

22

Q.

23

learning machine and the rest of that claim element.

24

A.

25

And let's turn now to the estimating parameters of a

Okay.
MR. NELSON:

Let me have slide 45, please.

Pazzani - direct

665

1

BY MR. NELSON:

2

Q.

3

illustrate?

4

A.

5

search box.

6

visited the Ram Trucks website and that is recorded in the

7

database.

8

Q.

9

there, too.

10

And can you explain what this slide is intending to

Yes.

Slide 45 shows a user typing a word into a
In this case, the word is "ram," and then they

And feel free to use the laser pointer you have up
I apologize.

The slides, it looks a little

small.

11

An let's go to the next portion of this

12

animation.

And can you explain what is happening here?

13

A.

14

visited a car repair website, and that is also stored in the

15

database.

16

Q.

And the next slide, please.

17

A.

Again, the user has typed another query, Mustang, got

18

a search result page back, found one of the results related

19

to what he was looking for, and went to the Mustang Club

20

website.

21

Q.

And next slide, please.

22

A.

Now the user has typed "Dodge" down to the Dodge.com

23

website and now over time we've continued to update the

24

database with each of the searches, each of the items the

25

user has clicked on.

Yes.

So now the user has typed "car repair" and

And that's the third entry in this database.

666

Pazzani - direct
1

Q.

And let me go to the next slide, please.

2

A.

Looked at used trucks and went to the Auto Trader

3

website.

4

Q.

5

depicted here now on slide 46?

6

A.

7

estimating, estimating the parameters of the learning

8

machine from the user-specific data.

9

about the searches the user has entered and the websites the

And next slide, please.

Yes.

And can you tell me what is

So this is showing a learning machine

The data is the data

10

user has visited, and we're going to put that data in the

11

learning machine and learn something about the interests of

12

the user.

13

Q.

14

machine represent?

15

A.

16

learning machine.

17

learning machine is going to become specific to the user.

18

When it analyzes the user-specific data, it will be a user

19

model.

And what is sort of a blank thing, a blank learning

Well, the blank learning machine is a generic

20
21

It has not learned yet.

MR. NELSON:

And then that

And can we run that animation,

please?

22

THE WITNESS:

So now this data goes into the

23

learning machine and we can see things like there's a topic,

24

cars.

25

BY MR. NELSON:

The topic "cars" has a high weight.

667

Pazzani - direct
1

Q.

Is that what the lights are intended to represent?

2

A.

Yes.

3

someone is very interested in cars.

4

doesn't have a weight.

5

websites there's an interest in sports.

6

weights that have a value between 1 and 7.

7

just be on, off.

8

an online shopper.

The topic cars has a high weight, indicating

9

The topic animals

He hasn't gone to any animal
And some of the
Others might be

We've also figured out this person is also

Now, we are depicting a model of him as

10

someone wearing a Nascar hat, but it's really the parameters

11

and the values that the system really uses to depict to

12

understand what the interests of the user are.

13

MR. NELSON:

14

please.

15

BY MR. NELSON:

16

Q.

17

learning machine that we just showed?

18

portion you referred to earlier?

19

A.

20

of the learning.

21

learning machine.

22

parameters.

23

to make predictions about other documents, to see whether

24

the user would be interested in them.

25

Go back here.

And let me have the next slide,
One back.

Sorry.

46.

And so what is the, what is the aspect of the

Yes.

Is that the training

So on the right is, again, the training portion
On the left is the training portion of the
That's the part that has learned the

And then we're going to use that model later on

MR. NELSON:

And let me have the next slide,

668

Pazzani - direct
1

please.

2

BY MR. NELSON:

3

Q.

4

slide, please.

5

A.

6

this user.

7

jaguar, and we get a pile of documents.

8

of these documents, we can estimate the probability the user

9

would be interested in that document based on model we've

10

And now let's go on to talk about step 38, the next

So here we have the user model that's specific to
We type a word, the user types something like
And then for each

learned about the user.

11

So we compare that model to the document.

12

it's looking at one of those green jaguar, the animal

13

documents, and it hopefully predicts a low probability that

14

the user would be interested in this.

15

Q.

And the next demonstration?

16

A.

On the other hand, one of the car-related Jaguar

17

documents is compared to the user model, and in this case,

18

the user, predicts the user would be interested in that.

19

There's a .89 probability.

20

the user things he's likely to be interested in.

21

Q.

22

on here?

23

A.

24

This is Brianne animals, if you like, and you can see that

25

the car light is low and the animal light is high.

So what we want to do is show

And next portion of it.

Yes.

Here,

Can you explain what's going

So this is showing a different user model.

So this

669

Pazzani - direct
1

is a user who is an animal lover caricature depicted as

2

someone as a model wearing a safari hat.
MR. NELSON:

3
4

Can you run the next animations

here?
THE WITNESS:

5

Oh, and in this case, can we go

6

back one?

7

document about the jaguar, the animal, and predicted there

8

would be a .92 probability that this user would be

9

interested in that, better than 50-percent chance it's

10

In this case, it has compared the model to the

likely they want to see this document.

11

MR. NELSON:

And let's go to the next slide,

12

please.

13

BY MR. NELSON:

14

Q.

15

slide.

16

A.

17

services.

18

We have a set of documents all about jaguars we can look at,

19

but one of them wants to see one type and the other wants to

20

see the other type.

21

the probability of interest and order the documents by the

22

ones they are most likely to click on.

And so let's now talk about step 40 and the next

Okay.

So this is talk talking about personalizing

So we have these two users with different models.

And we can use their model to predict

23

So we'll see that the man, who is the car lover,

24

gets lots of car documents, and the woman, who is the animal

25

lover, gets lots of animal documents.

Each of them contain

Pazzani - direct

670

1

the word "jaguar."

2

Q.

3

represent web pages?

4

A.

Yes, they are.

5

Q.

So let's turn, let's turn now to the next slide.

6

I want to talk a little bit about Google systems.

And these documents here, these are intended to

7

And

And, first of all, does Google use a lot of

8

different terminology that's, you know, kind of difficult to

9

understand until you understand the nomenclature?

10

A.

Yes.

11

a lot of unique Google playful terminology.

12

There's a lot of computer terminology and also

MR. NELSON:

And so let's talk about that a

13

little bit.

14

BY MR. NELSON:

15

Q.

16

profilers/profiles, and

17

A.

Okay.

18

Q.

Let's start with users.

19

to PTX-576, please?

20

A.

(Witness complies.)

21

Q.

Can you identify that document?

22

A.

Yes.

And I want to talk specifically about users, Kansas,

23
24
25

May I have the next slide?

as some of the Google terms.

Can you turn in your binder

Got it.

PTX-576 is a Google document entitled Gaia.
MR. NELSON:

And I'd like to offer PTX-576 into

evidence.
MR. VERHOEVEN:

No objection, Your Honor.

671

Pazzani - direct
1

THE COURT:

It's admitted.

2

(PTX-576 is admitted into evidence.)

3

MR. NELSON:

4

Can you further then pull up the pullout as

Can you put up Exhibit 576, please.

5

well?

6

BY MR. NELSON:

7

Q.

So what is this document tell you, say about Gaia?

8

A.

It says that Gaia is the user ID management system

9

used by all Google products.

It's used for signed-in users,

10

those that have accounts.

They created an account and a

11

password.

12

computer number called a Gaia ID that uniquely identifies

13

that user.

And then Google creates a unique 64-bit number, a

14

And I think earlier you saw associated with the

15
16
17

MR. NELSON:

Let me have PTX-12.

18

BY MR. NELSON:

19

Q.

And can you identify PTX-12?

20

A.

Yes.

21

remember

22

Q.

And is it a Google document?

23

A.

Yes, it's a Google document.

24
25

PTX-12 is called

You might

is actually a cookie.

MR. NELSON:
evidence.

.

Let me ask PTX-12 be admitted into

672

Pazzani - direct
1

MR. VERHOEVEN:

2

THE COURT:

3

(PTX-12 is admitted into evidence.)

4

Q.

6

It's admitted.

BY MR. NELSON:

5

No objection, Your Honor.

A.

What is the

7
8
9

MR. NELSON:

10

BY MR. NELSON:

12

Q.

And

the next portion.

11

Let me have slide 57, please.

13

And is this Google's document evidencing what a
document is?

14

MR. VERHOEVEN:

Objection.

Leading, Your Honor.

15

BY MR. NELSON:

16

Q.

What does this document tell you?

17

A.

This document tells how Google identifies

18

non-logged-in users.

19
20
21
22
23

Q.

And let me have you turn in your binder to PTX-113.

24

A.

(Witness complies.)

25

Q.

And can you identify that document?

673

Pazzani - direct
1

A.

Yes.

2

Quality.

3

Google uses in search ads.

4

Q.

Is this a Google document?

5

A.

Yes, it's a Google document.

6

presentation.

So it describes the personalization system that

MR. NELSON:

7
8

This is a document called User-Based Ads

A PowerPoint

I ask that PTX-13 be moved into

evidence.

9

MR. VERHOEVEN:

No objection.

10

THE COURT:

It's admitted.

11

(PTX-13 is admitted into evidence.)

12

MR. NELSON:

May I have slide 58, please.

And

13

the second portion.

14

BY MR. NELSON:

15

Q.

What does that document tell you?

16

A.

Well, it says that in the Search Ad system, a user is

17

defined by a prefID cookie.

18

cookies by different products in Google.

19

prefID cookie that Search Ads used to use and now Search Ads

20

uses then Gaia and

21

Q.

Let me have you turn to PTX-736, please.

22

A.

(Witness complies.)

23

Q.

Can you identify that document?

24

A.

Yes.

25

Ads or ID Ads.

So there are different types of
This is call the

IDs as well.

Yes.

This is a Google document called Interest-Based

674

Pazzani - direct
1

Q.

And you identify what is in that document?

2

Let me ask is it a Google document?

3

A.

Yes, this is a Google document.

4
5

MR. NELSON:

Let me ask that Exhibit 736 be

moved into evidence.

6

MR. VERHOEVEN:

7

THE COURT:

8

MR. NELSON:

10

THE COURT:

MR. NELSON:

13

THE COURT:

MR. NELSON:

16

THE COURT:

And

It's 113, Your Honor.
So you did not mean to offer 13 into

No, I did not.
Okay.

But you did mean to offer

Correct.

Thank you, Your Honor.

There is no objection to 113?

Correct.

18

MR. VERHOEVEN:

19

THE COURT:

20

736 is admitted.

113?

15

17

All right.

evidence, did you?

12

14

No objection.

the one on the screen, is it 113 or is it 13?

9

11

Oh, stop.

That's correct, Your Honor.

So 113 is admitted but not 13.

Go

ahead.

21

MR. VERHOEVEN:

Thank you, Your Honor.

22

(PTX-113 is admitted into evidence.)

23

MR. NELSON:

Can you put up slide 59, please.

24

And the next portion of it.

25

BY MR. NELSON:

675

Pazzani - direct
1

Q.

What is this document say?

2

A.

Well, this document describes a fourth way of

3

identifying users in Google.

4

testimony about this yesterday from Google witnesses.

5

is called the DoubleClick cookie and it identifies users in

6

the Content Ad system.

7

websites like LA Times or my blog.

8
9

MR. NELSON:

And you heard a lot of
This

Those are the ads that are on the

And let me turn to slide 60,

please.

10

BY MR. NELSON:

11

Q.

12

you to PTX-13 in your binder.

13

A.

Yes.

14

Q.

And can you tell me what PTX-13 is?

15

A.

Yes, PTX-13 is a Google document entitled Kansas.

So let's not talk about Kansas.

And if I can direct

16

MR. NELSON:

And I offer PTX-13 into evidence.

17

MR. VERHOEVEN:

18

THE COURT:

19

(PTX-13 is admitted into evidence.)

20

MR. NELSON:

No objection, Your Honor.

Okay.

13 is now admitted.

Can I have slide 61, please.

21

BY MR. NELSON:

22

Q.

And what is this document tell you?

23

A.

This document describes Kansas which is essentially

24

a large database system that Google uses to store user

25

information.

Pazzani - direct

676

1
Google says it's an infrastructure for storing

2
3

large amounts of user information and profiles for online

4

retrieval and personalization.

5

So this is really where the user model is.

We

6

think it's called Kansas because Kansas is in the center of

7

the country, central to Google.

8

Q.

9

use Kansas?

Can you identify which of the accused Google systems

10

A.

Yes.

11

Kansas to store the searches the users -- well, Search uses

12

Kansas to store the searches the users have done as well as

13

the results the users have seen and the results the users

14

clicked on.

15

I think all of them do.

So Search Ads uses

Search Ads uses it to look at the searches the

16

users have done again, the ads the user has seen and the ads

17

the user has clicked on.

18

Content Ads uses Kansas to store, again, the ads

19

the user has seen and the ads the user has clicked on and

20

also the web pages on which those ads occurred.

21

And that YouTube video ads also uses it.

22

Q.

Does Kansas also store timestamps?

23

A.

Yes.

24

are the timestamps.

25

Q.

So associated with each of those interactions

Let me turn to, or actually let me direct you in your

677

Pazzani - direct
1

binder to PTX-14.
Can you look at PTX-14?

2
3

A.

Yes.

4

Q.

What is that document?

5

A.

That document is the Google Developers Handbook.

6
7

MR. NELSON:

I'd like to move PTX-14 into

evidence.

8

MR. VERHOEVEN:

9

THE COURT:

No objection, Your Honor.

It's admitted.

10

(PTX-14 is admitted into evidence.)

11

MR. NELSON:

Can you put up slide 62, please?

12

And the next portion.

13

BY MR. NELSON:

14

Q.

What is this portion of the document refer to?

15

A.

This portion of the Google Developer's Handbook

16

further describes Kansas.

17

is that Kansas has a variety of user keys.

18

cookies that with discussed earlier, the Gaia ID, the

19

prefID, and the

20

Q.

Let me have you turn in your binder to PTX-15?

21

A.

(Witness complies.)

22

Q.

What is PTX-15?

23

A.

PTX-15 is entitled Data Available to Personalized

24

Search.

25

And what is most important here

ID.

Those are the

Also the mobile phone number.

It's a Google document.
MR. NELSON:

And let me offer PTX-15 into

Pazzani - direct
1

678

evidence.

2

MR. VERHOEVEN:

No objection.

3

THE COURT:

4

(PTX-15 is admitted into evidence.)

5

MR. NELSON:

It's admitted.

Can I have slide 63, please?

6

BY MR. NELSON:

7

Q.

8

left is intending to depict?

9

A.

And so can you tell me what the table here on the

Yes.

This depicts one of the tables in Kansas.

10

called a Gaia table.

11

It's

And that stores the information

associated with a Gaia ID, a user in Kansas.

12
13
14
15

Another important column in Kansas is

16
17
18
19
20
21
22

MR. NELSON:

23

this slide.

24

BY MR. NELSON:

25

Q.

Let me have the next portion of

So now, can you tell me what the prefID table refers

679

Pazzani - direct
1

to?

2

A.

3

prefID table is just associated with one of the other

4

cookies, the prefID cookie,

Yes.

This is very similar to the Gaia table.

The

5
6
7
8
9

Q.

And for search, the prefID isn't used any more; is

10

that right?

11

A.

12

non-signed-in users and Gaia for signed-in users.

13

Q.

14

Kansas?

15

A.

Yes.

16

Q.

And let me turn your attention in your binder to

17

PTX-395.

18

A.

(Witness complies.)

19

Q.

What is PTX-395?

20

A.

PTX-395 is a Google document entitled How UBAQ --

21

that is Users-Based Ad Quality -- and SmartAds Work

22

Together.

Yes, that's correct.

And does the

23
24
25

MR. NELSON:

It's just Zwieback for

store similar information in

I'd like to move PTX-395 into

evidence.
MR. VERHOEVEN:

No objection, Your Honor.

680

Pazzani - direct
1

THE COURT:

It's admitted.

2

(PTX-395 is admitted into evidence.)

3

MR. NELSON:

Can I have slide 65, please?

4

BY MR. NELSON:

5

Q.

What is this document now tell you about?

6

A.

This document describes some of the columns in Kansas

7

that store different types of data for search ads.

8
9
10
11
12
13
14

And during that

it can learn a lot

15

about you.

16

Q.

17

of keeping this information for

18

A.

19

the quality of ads presented to the user, to personalize

20

what the user sees so they're more likely to click on them.

And just like with the longer ones, what is the goal
?

The goal in keeping this information is to improve

21

MR. NELSON:

And let me turn to the next Google

22

terminology here.

May I have slide 66.

23

BY MR. NELSON:

24

Q.

And let's talk about profilers and profiles.

25

A.

Okay.

681

Pazzani - direct
1

Q.

Can you turn in your binder to PTX-213?

2

A.

(Witness complies.)

3

Q.

And can you tell me what PTX-213 is?

4

A.

PTX-213 is a Google document entitled Personalization

5

Profiles, Exercises and Tips.

6
7

MR. NELSON:

Yes.

I'd like to offer Exhibit PTX-213

into evidence.

8

MR. VERHOEVEN:

9

THE COURT:

No objection, Your Honor.

It's admitted.

10

(PTX-213 is admitted into evidence.)

11

MR. NELSON:

May I have slide 67, please?

12

BY MR. NELSON:

13

Q.

14

is intended to represent?

15

A.

16

in Google.

17

of generating profiles for users.

18

inferred information about a user's preferences.

19

profiles can then be used for a number of purposes, such as

20

twiddling search results, or making recommendations.

And can you tell me what this portion of Exhibit 213

This starts to describe the personalization process

21

In essence, personalization consists largely
That is some kind of
These

Twiddling search results is changing the order

22

of the search results so they're more personally relevant to

23

the user.

24
25

MR. NELSON:
please.

And let me have the next slide,

Pazzani - direct

682

1

BY MR. NELSON:

2

Q.

3

you tell me what this slide says?

4

A.

5

how Google generates a profile of the user and it says the

6

profiles could be the categories we think the users are

7

interested in, and then it says Google has a nice profiling

8

infrastructure that makes it easy to look at each user's

9

Kansas data, compute a profile from this data, and then

And this is part of the same exhibit, PTX-213.

Yes.

Can

This goes into a little bit more detail about

10

store that profile back into Kansas so it can be used to

11

personalize the search results of the user.

12

Q.

13

represented of our user, our car user, our animal user?

14

A.

15

the parameters.

16

Q.

And PTX-33 and 34 are those examples?

17

A.

Yes.

18

Q.

Let me turn to the next term here which is

19

Can you tell me what

20

A.

21

information about documents.

22

the web, for example.

23

Q.

Can you turn to Exhibit 16, please, in your binder?

24

A.

Yes.

25

Q.

Can you tell me what that document is?

And is that profile, is that the base that was

That is one way to represent it, but it's actually
That's the important part.

Yes.

.

is?
is Google system for storing
The documents that are out on

(Witness complies.)

Pazzani - direct
1

A.

Yes.

Exhibit 16 is a Google document entitled

Developers Guide.

2

MR. NELSON:

3
4

And I'd like to offer Exhibit 16

into evidence.

5

THE COURT:

6

MR. VERHOEVEN:

7

683

Any objection to 16?
Just considering it, Your Honor.

One second.

8

No objection to the document.

9

THE COURT:

10

Okay.

It's admitted.

(PTX-16 is admitted into evidence.)

11

BY MR. NELSON:

12

Q.

13

document.

14

A.

15

about the

16

says that the

17

are tokenized as well as indexed so that they can be browsed

18

as well as searched.

19

And let me direct your attention to a portion of the
Can you explain this document or this portion?

Yes.

So this goes into a little bit more detail
database, the repository, and it basically
repository is a set of documents that

So this tokenization process is finding all the

20

important words in this document and associating with that

21

document so later on when you are searching for something

22

like jaguar you can find the documents that contain jaguar.

23

Then it goes on to say that this, each document

24

in

has a global document ID that can be used to

25

specify a document in the whole repository.

Pazzani - direct

684

1
2
3
4
So it's not just like the address of an

5
6

envelope.

It's like you have opened up the envelope and

7

read all the important words in it, recorded it in the

8

database and then stored the address associated with those

9

important words so that you can find that envelope later on

10

by the words that are in it.

11

Q.

Let me direct your attention to PTX-372.

12

A.

(Witness complies.)

13

Q.

Can you identify that document?

14

A.

Yes.

15
16

Okay.

372 is the Google glossary.
MR. NELSON:

I'd like to offer PTX-372 into

evidence.

17

MR. VERHOEVEN:

No objection, Your Honor.

18

THE COURT:

19

(PTX-372 is admitted into evidence.)

It's admitted.

20

BY MR. NELSON:

21

Q.

22

document ID?

23

A.

24

It says each document that Google indices has a unique doc

25

ID assigned to it, generated from the URL of the document

Can you tell me what the Google glossary says about

Yes.

So the Google glossary defines the document ID.

Pazzani - direct
1

685

essentially.

2
3
4
5
6
MR. NELSON:

7
8

please.

9

And let me have the next slide,

BY MR. NELSON:

10

Q.

And so I want to shift gears a little bit now and

11

talk about the Google accused systems sort of in the context

12

of the terminology and the patents.
Let's talk about Google Search first and

13
14

personalization.

15

A.

Yes.

16

Q.

Can you please identify PTX-21?

17

A.

Yes.

18

Q.

And what is it?

19

A.

Oh.

20

Google.

It's an overview of personalization efforts at

It's a Google document.

21
22

Can you turn to PTX-21, please?

MR. NELSON:

I'd like to move PTX-21 into

evidence.

23

MR. VERHOEVEN:

No objection, Your Honor.

24

THE COURT:

25

(PTX-21 is admitted into evidence.)

It's admitted.

Pazzani - direct
1

MR. NELSON:

686

The next slide, please.

2

BY MR. NELSON:

3

Q.

Can you tell me what the pullouts say?

4

A.

Yes.

5

at Google.

6

to improve search quality based on what we know about a

7

user.

8

in personalized search quality.

9

searches and click history.

This discusses the search personalization team
The goal of the search personalization team is

And then it goes on to say that they have advantages
They store a user's

That is the Project Kansas that

10

we talked about.

11

search results based on this search history and click

12

history.

13

Q.

14

detail here probably later today and this afternoon.

15

we'll move to -- can you look at PTX-22, please?

16

A.

Yes.

17

Q.

What is PTX-22?

18

A.

PTX-22 is a PowerPoint presentation entitled

19

Personalized Search.

Now, I know we'll talk about those in a lot more

20
21

Then they have algorithms for customizing

It's a Google document.

MR. NELSON:

I'd like to move PTX-22 into

evidence.

22

MR. VERHOEVEN:

23

THE COURT:

24

(PTX-22 is admitted into evidence.)

25

BY MR. NELSON:

No objection.

It's admitted.

So

687

Pazzani - direct
1

Q.

Dr. Pazzani, there is a drawing in PTX-22, and rather

2

than put it on the board here, we blew it up so it would be

3

a little bit earlier to see and I would ask you come down

4

and walk through a sort of overview of the Google Search

5

system using that drawing and I'll put it up here.

6

A.

Okay.

7

Q.

Can you just walk through sort of the general

8

description of the Google Search system using this figure

9

from PTX-22?

10

A.

Yes.

11

search system in Google.

12

user types a query to the Google web server and the user

13

gets a response, search results back that you can click on.

14

So that is the part that is visible to the user.

15
16

The part that the user sees is the

And then this part is all behind the scenes what
is happening at Google.

17
18

So this describes the personalization and

So some of the things that are happening is when
you type the query and you signed into Google, associate

19
20
21

Q.

What happens next?

22

A.

Well, let's describe a couple things.

23

happens is after you typed a number of queries and clicked

24

on a number of results, Google can learn a profile of your

25

interests.

Now, what

So this system called a profiler looks at your

688

Pazzani - direct
1

search history in Kansas and learns the interests.
-- I'm sorry.

2
3
4

then stores that profile back into Kansas.

5

It knows the documents are about animals and

6

cars

7
8
9

Once it has your profile, when you type a search

10

result, that search result, the query plus your profile go

11

to

12

you based on your profile.

13

to the web server and then sent to the user.

14

Q.

15

level, the portions of this drawing that would be the

16

training portion of the learning machines?

17

A.

, and the search results are reordered for
Some of them are

, given

And can you identify the portions of just on a high

Sure.

18

MR. VERHOEVEN:

19

outside the scope of the expert report.

20
21
22
23

THE COURT:

Your Honor, I'm objecting

All right.

What is your response,

Mr. Nelson?
MR. NELSON:

This is totally in the expert

report, in numerous paragraphs.

24

Jennifer.

25

THE COURT:

You will have to identify it more.

689

Pazzani - direct
1

(Pause.)

2

MR. NELSON:

3

Honor.

I'm sorry.

4

The figure is in paragraph 53, Your

The document is in paragraph 53.

The analysis is throughout the report but in

5

particular the portion that is going to talk about now

6

begins at paragraph 163 and ends at paragraph 197.

7

THE COURT:

8

MR. NELSON:

9

THE COURT:

163 to 197?
Yes.
Mr. Verhoeven, you can review

10

that quickly and let me know if you still have your

11

objection.

12

MR. VERHOEVEN:

13

maintain the objection.

14

Yes, your Honor.

We still

about something different.

15
16

THE COURT:

We believe this section is talking

Tell me again what the question is

that you've asked the witness.

17

MR. NELSON:

Well, the claim elements have

18

estimating parameters of a learning machine, using those

19

parameters to create a user model specific to the user.

20

Element C in the claims.

21

level way to describe that, but within the words of the

22

report --

23

THE COURT:

24

MR. NELSON:

25

This figure is sort of a high

What's the question that -Oh, the question was:

Can you

identify the portion that would be the training aspect of

690

Pazzani - direct
1

the learning machine or user model?

2

THE COURT:

All right.

And you are saying that

3

the training aspect of the learning machine is discussed in

4

these 30 paragraphs of the report?

5

MR. NELSON:

The objection is to the word

6

"training," maybe not, but the training aspect of it is

7

discussed elsewhere in the report.

8

the portions of the learning machine that he identified is

9

discussed in these paragraphs of the report.

10
11

THE COURT:

This is the, these are

Mr. Verhoeven, is that responsive to

your objection?

12

MR. VERHOEVEN:

I guess my objection, Your

13

Honor, we read it in the report, had the expert go through

14

this document and start describing all the different

15

portions of the system.

16

right there I don't believe is anywhere in the reference

17

section, Your Honor.

18

And, in particular, that question

If counsel would like to stay within the

19

description in that section, we would have no objection, but

20

we feel like they're going beyond the scope of the expert

21

report.

22
23
24
25

THE COURT:

All right.

Do you want to try to

restate it or do you want me to rule on the objection?
MR. NELSON:
BY MR. NELSON:

Well, we disagree.

I can ask it.

691

Pazzani - direct
1

Q.

Can you just describe using this document generally

2

how the system learns and then applies what is learned?

3
4

MR. VERHOEVEN:

Again, Your Honor, there's no

disclosure.

5

THE COURT:

All right.

Well, ladies and

6

gentlemen of the jury, we're going to take our break a

7

little bit early this morning.

8

need to attend to with counsel, so no talking about the case

9

during the break, and we'll get you back here just as soon

10

There are some matters I

as we can.

11

(The jury was excused for a short recess.)

12

THE COURT:

All right.

So, Mr. Nelson, it's 163

13

and 193 I need to review in order to rule on this objection.

14

Correct?

15

MR. NELSON:

I believe so, yes.

And then also,

16

Your Honor, the second portion of this is the twiddling

17

portion with Kansas, that would be Element D, and that

18

section, this would be for stuff coming up here down the

19

road.

20

208 through 245.

21

That section for Element E and F would be paragraphs

And what he does in the report is he walks

22

through and identifies what is the learning machine, what is

23

the learning machine and then how it's applied.

24

training portion is Kansas plus the profilers, and the

25

applying portion is Kansas plus the profilers plus the

And so the

692

Pazzani - direct
1
2

Kaltix twiddler which is doing the applications.
Also for background, there's background in here

3

in paragraphs 62 to 96 generally on all of this stuff as

4

well, and this figure is used.

5

THE COURT:

6

MR. NELSON:

All right.
The report is very comprehensive

7

and we thought an easier way to present it was to just sort

8

of walk through this figure and that's what we're trying to

9

do, which it is in the report.

10
11
12

THE COURT:

Mr. Verhoeven, anything else I

should have in mind as I review your objection?
MR. VERHOEVEN:

Yes, yes.

When you read through

13

the passages, it's very unclear what portion of the system

14

the expert witness is pointing to.

15

use this figure and point to these portions in the system in

16

our view, Your Honor.

And we certainly did not

17

So this is a surprise to us.

18

we read the expert report, were uncertain as to these

19

aspects and so for us, it's a surprise the a least.

20

the way we feel.

21

THE COURT:

22

MR. NELSON:

We actually, when

That's

Okay.
They certainly explored this during

23

Dr. Pazzani's deposition as well and can address some of the

24

issues on cross as well if they feel that they didn't know

25

what was accused.

693

Pazzani - direct
1
2

THE COURT:

And you said earlier, Mr. Nelson, I

could find this figure in the report?

3

MR. NELSON:

Paragraph 55.

Correct, Jennifer?

4

Where is this figure?

5

I'm trying to find where the figure is exactly.

6

It's Exhibit 17 to thinks report.

THE COURT:

Mr. Verhoeven, is there a dispute as

7

to whether Exhibit 17 to the report contains the figure that

8

we're looking at?

9
10

I don't have the exhibit.

MS. BENNETT:

Exhibit 17 is in paragraph 53 of

his report.

11

MR. NELSON:

Paragraph 53.

12

appear on that page, but the paragraph.

13

figure is in here, too.

14

The figure does not
I believe the

where.

I'm just trying to find out exactly

15

It is footnote 24 that's in his description of

16

personalized search and that is Exhibit 17 and I think the

17

figure is reproduced later.

18

THE COURT:

Right.

So you're telling me that

19

the figure, which we now have as a demonstrative and which

20

is labeled PTX-22, it looks like from here?

21

MR. VERHOEVEN:

22

MR. NELSON:

23
24
25

admitted 22.

It's an exhibit, Your Honor.

It's an exhibit.

It's a portion of

I think it's like page 6 of it.
THE COURT:

You are saying if I have page 17 to

Pazzani's report, it would be the same thing.

694

Pazzani - direct
1

MR. NELSON:

Correct.

2

THE COURT:

3

MR. VERHOEVEN:

Do you agree with that?
I don't know.

We'll take his represent.

We're checking

4

the 17 part.

Currently, if you

5

look at what he cited, we're in a different section.

6

in paragraph 53.

7

many, many exhibits, Your Honor.

8

has pointed to is a footnote that does not explain the

9

document, and it's a different section than what he's

We're

These reports are very voluminous and
And the only reference he

10

talking about here that he's citing you to that has the

11

substantive description.

12

And so there's no way we could divine that he

13

was going to talk about this specific exhibit because just

14

from the fact that it was cited in footnote 23 in a

15

different section of the report from the substance section

16

he's going into now.

17
18

THE COURT:
look at it.

All right.

Well, we'll go take a

We'll be in recess.

19
20
21

(Proceedings resumed after the short recess.)

22

THE COURT:

Have a seat for a moment.

I

23

reviewed, to the extent we could, the materials that were

24

referenced, and pending Google's objection to the testimony

25

of Dr. Pazzani as beyond the scope of the expert report,

Pazzani - direct
1
2

695

Google's objection is overruled.
It appears that the witness is going to testify

3

and maybe even has been testifying at somewhat a higher

4

level of generality about those aspects of Google that

5

constitute a learning machine, higher level of generality

6

than appears in the expert report.

7

specifics in the context of specific user profiles.

8
9

The report details

The generality offered here in court, while not
literally in the report, is helpful, I think to the jury,

10

should not be a surprise to Google, and which I believe has

11

seen the demonstrative and has seen the exhibits that were

12

going to be used on direct.

13

Court's view, unfairly prejudicial to Google.

And the context is not, in the

14

The Court is going to charge Google five minutes

15

for the argument that we devoted to the objection while here

16

in Court, plus whatever time it's taking me now to

17

articulate my ruling, but I will not charge Google or

18

anybody for the 15 minutes or so it took us to figure this

19

all out back in chambers, since it happened to come up

20

during the time when we were about to take a break anyway.

21

So that's the ruling.

22

MR. NELSON:

23

THE COURT:

24

MR. VERHOEVEN:

25

THE COURT:

Any questions about that?

No, Your Honor.
All right.

Any questions?

No, Your Honor.

All right.

696

Pazzani - direct
1

MR. VERHOEVEN:

2

THE COURT:

3

(The jury entered the courtroom.)

4

THE COURT:

5

Thank you.

We will bring the jury back in.

Welcome back, ladies and gentlemen.

We are ready to continue.

6

Mr. Nelson, you may proceed.

It has gotten very

7

cold in here now, predictably.

8

I'm sure that is true.

9

to see if we can get the temperature raised a little bit.

10

It's warm back there, yes,

It's an old building.

We're going

We'll see how that goes.

11

MR. NELSON:

12

Dr. Pazzani, may I ask you to come back down and

13

Thank you, your Honor.

continue your explanation, please?

14

THE WITNESS:

Yes.

15

(The witness left the witness stand and

16

approached the easel.)

17

BY MR. NELSON:

18

Q.

19

the training portions of the learning machine from this

20

drawing and Google.

21

A.

22

really the profiler.

So I think before the break I asked you to identify

Yes.

Can you do that, please?

The training portion of the learning machine is

23
24
25

And it estimates the user's interest in

697

Pazzani - direct
1

categories.

These are parameters, things like topics

2

like cars or animals, and that's how the profile is

3

represented.

4

Kansas.

5

Q.

6

profiles?

7

A.

8

They just stick around for about

9

Q.

And then the profile is stored back into

The long term profiles.

What about short-term

The short-term profiles are kept in the main memory.
.

And we're going to talk about this in a whole more

10

detail in a little while.

11

the Google Search engine works at a high level just walking

12

through this?

13

A.

14

user interactions and then the Kansas database is where the

15

interactions are updated to and associated with the Gaia ID

16

in this case.

Yes.

17

Can you continue explaining how

So it's the Google web server that monitors the

And then the profile goes to the Google web

18

server and the twiddler.

And the profile is the user's

19

interest.

20

the user, and those that have the --

It's compared to the possible documents to show

21
22
23
24
25

And we'll get into a
little about math later.

They get

.

So the search results are individualized by
the users and the twiddlers and then they are sent back to

698

Pazzani - direct
1

the user, and then you can monitor what the user clicks on

2

to see if you've got it right, and if they click on the top

3

things, you have a good model of the user.

4

click on, you go to the next page and you get to learn

5

something about the user.

6

Q.

7

profiles, is that the phase from our graphics?

8

A.

9

it's really the weights.

If they don't

And just to tie it back to the graphics, the

That's how we are representing that as a phase, but

10

Q.

It's really a bunch of data?

11

A.

Yes.

12

Q.

All right.

Thank you.

13

(The witness resumed the witness stand.)

14

MR. NELSON:

15

THE COURT:

16

(Mr. Nelson removed the board from the easel.)

17

MR. NELSON:

May I move it?
You may move it.

May I have slide 75, please?

18

BY MR. NELSON:

19

Q.

20

determine its properties.

So let's talk now about analyzing a document to

21
22

MR. NELSON:

And may I approach the witness,

Your Honor?

23

THE COURT:

24

(Mr. Nelson handed an exhibit to the witness.)

25

BY MR. NELSON:

You may.

Pazzani - direct

699

1

Q.

Let me hand you something that's not in your binder.

2

A.

Okay.

3

Q.

And can you tell me what is Exhibit 202, please?

4

A.

Yes.

5

classification and personalized applications."

6
7

It is a Google document entitled "Text

MR. NELSON:

I'd like to move Exhibit 202 into

evidence, please.

8

MR. VERHOEVEN:

9

THE COURT:

10

No objection, Your Honor.

It's admitted.

(PTX-202 was admitted into evidence.)

11

MR. NELSON:

May I have slide 76, please.

12

BY MR. NELSON:

13

Q.

14

document discusses?

15

A.

16

classification, ways of analyzing text by classifying it

17

into topics.

18

open directory project classifier that was mentioned

19

earlier in the patent.

20

is a library that classifies text into selected ODP

21

categories.

22

Q.

23

document classification looks like?

24

A.

Yes, it did.

25

Q.

And can you turn to PTX-25, please.

And can you tell me about what this portion of the

Yes.

So in general, the document discusses text

And this describes the ODP classifier, the

And it says that the ODP model

And did Google provide some examples of what the

Pazzani - direct
1

A.

Yes.

2

Q.

And please identify that.

3

A.

700

It is a Google document entitled

4
MR. NELSON:

5
6

I'd like to move PTX-25 into

evidence.

7

MR. VERHOEVEN:

8

THE COURT:

9

No objection, Your Honor.

It is admitted.

(PTX-25 was admitted into evidence.)

10

MR. NELSON:

May I have slide 77?

11

BY MR. NELSON:

12

Q.

13

the right?

14

A.

Yes.

15

page.

And the web page is http.MassEffect.biowear.com.

16

is about a computer game, essentially.

17

Q.

18

that's not part of the document, is it?

19

A.

No, it's not.

20

Q.

Yes.

21

A.

Well, the graphic is what is on that web page today.

22

I don't know what was on that page ten years ago.

23

Q.

24

exhibit is showing?

25

A.

And can you identify, first of all, the graphic on

The graphic is a, today's contents of a web
It

And just to make clear, the graphic on the right,

I mean, what is it intended to represent?

Can you look at PTX-25 and just describe what the

Yes.

The exhibit is showing the ODP categories that

Pazzani - direct

701

1

are used by the link profiler.

So each one of those lines

2

is a category and a hierarchy, and these are things like,

3

one of the things that is highlighted is Dungeon and Dragon

4

games.

5

classified it as a Dungeon and Dragon-type game.

So this is a computer game web page and it has

Now, we saw in the patent that documents can

6
7

have multiple topics, and it has actually identified about

8

ten topics in this document.

9

related to computer games.

10

Q.

11

And this says link.

They're all very closely

What is link?

A.

12
13
14
15
16

MR. NELSON:

And let me have the next slide,

17

please.

This is another portion of PTX-25.

18

BY MR. NELSON:

19

Q.

What does this one say?

20

A.

So this is the same web page, but it's a different

21

form of analysis.

22

Google,

This type of analysis is called dilip in

23
24
25

So this shows that, for instance, the computer
game website is associated with other websites, such as

702

Pazzani - direct
1

XBox.com.

2
MR. NELSON:

3

And let me have the next slide,

4

please.

This is slide 79.

This is another portion of

5

PTX-25.

6

BY MR. NELSON:

7

Q.

What does this slide depict?

8

A.

This slide depicts the same website with a third

9

categorization system called rephil.

Actually, the first

10

one was called phil and they did it again and called it

11

rephil.

12

And this also shows that the topics are

13

things like video games and XBox associated with this

14

particular page.

15

hierarchies, three different alternative ways of classifying

16

the world.

17

Q.

18

talk about the monitor user interactions with data.

19

have the next slide, please?

Let's turn to the next slide, please.

20
showing?

22

A.

24
25

And let's now
May I

And can you just explain what this slide is

21

23

But it's, if you like three different

Yes.

703

Pazzani - direct
1
2
3
4
5
6
7

Q.

This was part of your experiments that was done on

8

Google Systems?

9

A.

That's correct.

10

MR. NELSON:

Let me have the next slide, please.

11

BY MR. NELSON:

12

Q.

13

files.

14

the 1(a) and 1(b) and 1(d) are appearing.

15

intended to represent?

16

A.

17

the first claim of the '040 patent, but, again, my analysis

18

is not just by the words in that box, but it's by the entire

19

claim element.

20

Q.

21

explain what is going on here?

22

A.

So let's talk next about the user-specific data
And let me just ask a quick question.

You see where

What are those

Those are intended to represent the claim elements of

And let me have slide 83, please.

And can you

23
24
25

you can see
right here.

So that's that number there that gets put into

Pazzani - direct
1

the database.

2
3

704

So that's how you distinguish
.

There's a different ID associated with

that.

4

Then, it shows the query.

5
6
7
8
9
10

So, for instance, the hummingbird.net page was in position 5

11

and she clicked on it at this time,

12
13
14
15

Q.

16

represents?

17

A.

18

Internet.

19

that's one way of associating the document with the user.

20

The user has an ID.

21

document ID in the database column associated with the user

22

and then you can tell the user accessed that document on

23

that day.

24
25

And the

, the URLs

The URLs are the location of the document on the
That's the URL you would type in to get there and

The document has an ID.

MR. NELSON:
please.

what do those

You have the

And let me have the next slide,

705

Pazzani - direct
1

BY MR. NELSON:

2

Q.

3

learning machine element again.

4

A.

And let's talk about the estimating parameters of a

Okay.
MR. NELSON:

5
6

BY MR. NELSON:

7

Q.

8

this is?

9

A.

Let me have the next slide.

10

And this is back to our graphics.

Can you say what

Yes this is the abstract learning machine.

It's a

mathematical function or model.
MR. NELSON:

11

And can I have slide 86, please?

12

BY MR. NELSON:

13

Q.

And can you explain what is going on here?

14

A.

Yes.

15

estimated the parameters.

So this showing the learning machine that

16

Can I ask you to zoom in on this portion?

17

So what this shows are the

18

associated with

.

19
20
21
22
23
24
25

.

Pazzani - direct
Can we zoom out so we can see that on the

1
2

learning machine?
Now we represent that in the learning machine by

3
4

these numbers.
and the lights light up indicating the strength of that

5
6

association.
And again, we're depicting that with a woman

7
8

706

wearing a safari hat.

9
10
11

Q.

And do those change over time as the user does other

12

service?

13

MR. VERHOEVEN:

14

THE COURT:

Objection, leading.

Overruled.

15

BY MR. NELSON:

16

Q.

Go ahead.

17

A.

Yes.

18

and more accurate model.

19

pages, you don't really know what someone's interests are,

20

but after they have seen 50 or 100 web pages, then you know

21

much more.

22

that they don't really see the effect of personalization

23

until after about the first month or so.

24
25

The whole goal of learning is to create a more
So if you have only seen five web

In fact, there are Google documents that show

MR. NELSON:
BY MR. NELSON:

Let me have the next slide, please.

707

Pazzani - direct
1

Q.

2

A.

3

told you there are three different hierarchies or three

4

different categorization schemes.
of

.

And this

shows, for instance,
And we are representing this in the learning

7
8

I

This is the

5
6

And what is this slide intending to represent?

machine by lights that are either on or off.
There is not

9
10

that finer gradation that we saw

.

11

Q.

12

how are those determined whether or not they get in or out

13

of the profile?

14

A.

15

later,

And how are those categories represented as squares,

So I'm going to go into a lot of detail of that

16
17
18
19
20
21
22
23

MR. NELSON:

24

actually, go ahead and take it down.

25

BY MR. NELSON:

Let me have the next slide, please.

Pazzani - direct

708

1

Q.

Turn to your PTX-38, please.

2

A.

Yes.

3

Q.

What is that document?

4

A.

It is a document interested Personalized Search.

5

It's a Google document.

6

MR. NELSON:

7

I'd like to move PTX-38 into

evidence.

8

MR. VERHOEVEN:

9

THE COURT:

10

No objection.

It's admitted.

(PTX-38 is admitted into evidence.)

11

BY MR. NELSON:

12

Q.

Can you describe PTX-38?

13

A.

Yes.

14

discusses Personalized Search.

15

PTX-38 is a PowerPoint presentation that

MR. NELSON:

Can I have the pullout.

16

BY MR. NELSON:

17

Q.

What does this show?

18

A.

This shows a

19

Google learns over the course of

20

this case, they know that the user session contains the

21

query Boston.

22

website, the City of Boston.

23

that the topics of interest are Massachusetts, Boston, arts

24

and entertainment in Boston which that website describes.

25

Q.

.

So this is a profile
.

In

You typed Boston and then you clicked on a
And essentially it has learned

And if a user continued to search over that

Pazzani - direct

709

1

window, what would happen?

2

A.

3

enter into that.

4

Boston Celtics and then the sports categories would be

5

entered.

6

Boston and the animal categories would show up in there.

It would get better and better, more categories would
So some users might be interested in the

Others might be more interested in the aquarium in

7

MR. NELSON:

8

And let's talk about estimating the probability

9

May I have Slide 89, please.

that a user is interested in a document step.

10

May I have slide 90, please.

11

BY MR. NELSON:

12

Q.

13

depict?

14

A.

15

how one estimates the probability that the user is

16

interested in the document.

17
18
19
20
21
22
23
24
25

So can you explain what the graphic is intended to

Yes.

This graphic is depicting just to illustrate

So we have a user model that is specific to the
user.

710

Pazzani - direct
1

Q.

Does this one animate or not?

2

A.

No, this is just showing that in this case, the

5

Q.

And why is that?

6

A.

Yes, it is.

7

Q.

And why is that?

8

A.

Well, it's not precise.

9

learned a little bit about

3
4
Is that number an estimation?

First of all, we have only

10
11

We have only seen a few documents so

12

far.

13

Q.

14

result?

15

A.

16

with 100 percent accuracy, predict a user's interest.

What happens if you see more documents?

What is the

It becomes more and more accurate, but you can never,

17

MR. NELSON:

18

please.

19

BY MR. NELSON:

20

Q.

Let me have you turn to PTX-729, please.

21

A.

(Witness complies.)

22

Q.

Please identify Exhibit PTX-729?

23

A.

It is a Google document entitled Optimization For

24

User Profile

25

Whoops.

And let me have the next slide,

Stop.

MR. NELSON:

I'd like to move Exhibit 729 into

711

Pazzani - direct
1

evidence.

2

MR. VERHOEVEN:

3

THE COURT:

4

(PTX-729 is admitted into evidence.)

5

MR. NELSON:

6

Q.

8

It's admitted.

Let's put up slide 91.

BY MR. NELSON:

7

No objection, Your Honor.

A.

Can you tell me what is

?

9
10
11
12
13
14
15
16
17
18
19
20
21

But essentially what this is saying is Google

22

uses a number of

to predict the probability the

23

user will click on that.

Some of these

24

the Kansas database that are individualized to the user and

25

some of these are aggregate across all.

come from

In this case, we

712

Pazzani - direct
1

are only concentrating on these

2

that come from

Kansas.

3

MR. NELSON:

4

so let's talk now about using that estimating

5

Let me have slide 92, please.

probability to provide the personalized service.

6

Can I have slide 93?

7

BY MR. NELSON:

8

Q.

Can you tell me what this graphic is depicting?

9

A.

Yes.

So slide 93 is depicting from all of the

10

possible search results that contain the word "jaguar" which

11

ones

12

more of the animal documents and a couple other ones, music

13

and computers as well.

would be most interested in, and it has

14

But in general, what it has done is it has found

15

documents that are potentially of interest to

16

estimating her probability of interest and

17

.

18

.

by
them or

So they're at the

So, for instance, that

doesn't have to

19

scroll down to find them, they're within the

on

20

the screen.

21

Q.

Can you turn in your binder to PTX-41, please?

22

A.

(Witness complies.)

23

Q.

Please identify that document?

24

A.

It is a Google document entitled Twiddler Quick Start

25

Guide.

713

Pazzani - direct
MR. NELSON:

1
2

I'd like to move Exhibit 41 into

evidence.

3

THE COURT:

4

MR. VERHOEVEN:

5

Any objection?
No objection.

I'm sorry, Your

Honor.

6

THE COURT:

7

(PTX-41 is admitted into evidence.)

8

MR. NELSON:

9

It's admitted.

Can you put up slide 94, please.

BY MR. NELSON:

10

Q.

11

What is a twiddler?

A.

12
13
14
15
16

MR. NELSON:

And let me put up slide 95.

17

BY MR. NELSON:

18

Q.

What did Mr. Taher Haveliwala say about twiddler?

19

A.

So you saw this on video yesterday.

20

he said:

21

"Question:

22

He says:

23

"Answer:

24
25

But when asked,

And what is a twiddle?"

714

Pazzani - direct
1
2
3

Q.

And can I have you turn to PTX-44, please?

4

A.

Yes.

5

Q.

What is PTX-44?

6

A.

It is a Google document entitled Twiddler Catalog.

7
8

MR. NELSON:

I'd like to move PTX-44 into

evidence.

9

THE COURT:

Any objection to 44?

10

MR. VERHOEVEN:

11

THE COURT:

12

(PTX-44 is admitted into evidence.)

13

MR. NELSON:

14

Q.

16

A.

Please put up slide 96.

Twiddler?

17

It is admitted.

BY MR. NELSON:

15

No, Your Honor.

What does Google say about -- what is the Kaltix

The Kaltix Twiddler is the

18
19

that I have sketched and I
will describe in more detail later.

20

And this

21

is a user who at that time has opted into Personalized

22

Search.

23

Q.

Where did the name Kaltix come from?

24

A.

Kaltix was the company that Google acquired I think

25

in 2003.

The witness that we just saw and some others were

Pazzani - direct

715

1

founders of Kaltix, and then they became Google employees

2

and started the Kaltix Project to personalize search

3

results.

4

BY MR. NELSON:

5

Q.

Let me have you look at PTX-43.

6

A.

It's a Google could entitled Personalized Search For

7

Everyone.

8
9

MR. NELSON:

I'd like to move Exhibit 43 into

evidence.

10

MR. VERHOEVEN:

No objection.

11

THE COURT:

12

(PTX-43 is admitted into evidence.)

13

MR. NELSON:

It's admitted.

Can you put up the next slide.

14

BY MR. NELSON:

15

Q.

16

everyone?

17

A.

18

helping people get better search results by extending

19

personalized search to signed-out users worldwide.

What does Google say about personalization for

Well, it says that on December 4th, 2009, we're

20

Now when you search using Google, we'll be able

21

to better provide you with the most relevant results

22

possible.

23

Previously, we only offered personalized search

24

for signed-in users and only when they had web history

25

enabled.

What we're doing now is expanding personalized

716

Pazzani - direct
1

search so we can provide it to signed-out users as well.

2

Essentially, all users of Google whether they have signed in

3

or signed out.

4

Q.

5

signed-out users?

6

A.

7

we described a little earlier.

Is there a specific ID that is associated with the

The signed-out users use the

MR. NELSON:

8
9

cookie, the one

Let me have slide 99, please.

BY MR. NELSON:

10

Q.

Can you explain what is going on here?

11

A.

Yes.

12

insight.

So slide 99 it's a Google Search for the word

13

That's what it says up here.

14

Can you zoom in on that, actually?

15

So you can see this is a search done when

16

.

17

And now if you zoom back out?

18

You can see that the first search result is for

19

a Honda Insight.

20
21

That is a car.

The other search results I didn't know, but
there is insight communication, the word insight, et cetera.

22

MR. NELSON:

23

BY THE WITNESS:

24

A.

And could I have slide 100, please?

25

Slide 100 is the same search "insight" for their
accounts.

See, it's a little longer.

We

717

Pazzani - direct
1

won't zoom in but take my word for it.

2
3

That is

, and we'll see the Honda Insight is not the first
search results.

4

Why don't we zoom in on that?

5

So now the insight communications and other

6

insight things are the first things, and the Insight car I

7

think is the sixth or seventh result.

8

different Google users get different search results based on

9

what Google has learned about them.

10

So it shows that

It has personalized the

results to them.

11

MR. NELSON:

May I have the next slide, please.

12

BY MR. NELSON:

13

Q.

14

high level.

15

opportunity to stretch your legs a little bit.

16

A.

Okay.

17

Q.

Can you just, using this board, just sort of at a

18

high level walk through the functionality of the Google

19

Search Ads system explaining the different pieces of a

20

picture as appropriate?

21

A.

22

typing the search query.

23

repair, goes to the Google front end, the Google web server,

24

the balancer.

25

So let's generally talk about Google Search Ads at a
Let me set up the board to give you an

Thanks.

Sure.

So here we have the user.

That's the person

The search query, a word like car

The most important thing here is the ad mixer.

718

Pazzani - direct
1

That was described in some of the testimony you saw

2

yesterday.

3

related to the search query.

4

that an advertiser has said, display these when the words

5

"car repair" or "auto repair" are typed.

6

What the ad mixer does is it retrieves ads
These are ads, for instance,

And then Google wants to decide which ads to

7

display, and it uses a system called SmartAds as depicted

8

here by a wizard with a crystal ball, and that wizard with

9

the crystal ball is trying to predict which ads the user is

10

most likely to click on.

11

calculate, they predict.

12

It's an estimation.

Wizards don't

SmartAds uses the UBAQ profile, the User-Based

13

Ad Quality profile, we'll see that in a little bit more

14

detail, which contains information that Google has learned

15

about the user, what type of ads they click on, what type of

16

ads they ignore.

17

After we predict here what ads the user is most

18

likely to click on, the ad mixer conducts an auction.

It

19

combines how likely the user is to click on an ad with how

20

much the advertiser is willing to pay.

21

ads that maximize Google's revenue, a combination of your

22

likely to click on it but you will not pay on it or you are

23

less likely to click on it and the advertiser is going to

24

pay a lot.

25

Google the most revenue.

And it selects the

It balances that and figures out which will earn

719

Pazzani - direct
1

There is a couple other factors that go into it.

2

The ads go into the Google web server, the Google front end,

3

and then the user gets the results page.

4

search results plus the ads that are on the right side.

5

THE COURT:

6

THE WITNESS:

7

Thank you.

Thank you.

Oh, I think you are

going to ask me to come back.

8
9

All right.

Those are the

MR. NELSON:

I will.

I need you to look at

PTX-115 first.

10

THE WITNESS:

Okay.

(Resuming witness stand.)

11

BY MR. NELSON:

12

Q.

Can you identify PTX-115, please?

13

A.

Yes.

14

Quality System and Team Overview.

15
16

PTX-115 is a Google document entitled Ads

MR. NELSON:
board?

17

THE COURT:

18

MR. NELSON:

19

Can I get the next slide up on the

Do you want to offer that document?
Yes, I do, please.

I offer PTX-115

into evidence.

20

MR. VERHOEVEN:

No objection.

21

THE COURT:

22

(PTX-115 is admitted into evidence.)

It's admitted.

23

BY MR. NELSON:

24

Q.

Can you explain what this document is?

25

A.

Yes.

This document describes the ads quality, the

Pazzani - direct

720

1

system that tries to improve the quality of ads shown to a

2

user.

3

interested in them, they're more likely to click on them.

Basically, users are more -- if they're more

4

MR. NELSON:

And can I get you to zoom in a

5

little bit on the portion right there that Dr. Pazzani is

6

pointing to that says "UBAQ?"

7

And can you explain what UBAQ and UBAQ adjusted

8

pCTRs are?

9

A.

10
11
12
13
14
15

MR. NELSON:

Could I get the portion blown up in

16

the next part of the slide?

I think it's the next.

Oh,

17

there we go.

18

BY MR. NELSON:

19

Q.

20

some of these so I'm not sure we need to go back through

21

them.

Let me have you turn in your notebook to PTX-222.

22

A.

Okay.

23

Q.

And please identify PTX-222?

24

A.

It is a Google document entitled CTR Prediction in

25

Content Ads, AdSense For Content.

And just at a high level -- we already talked about

CTR stands for

721

Pazzani - direct
1

click-through rate.

2

user is going to click on it or not.

3

MR. NELSON:

It's trying to estimate whether the

I think that has already been

4

offered into evidence.

5

THE COURT:

6

MR. VERHOEVEN:

7

THE COURT:

8

(PTX-222 is admitted into evidence.)

9

If not, I'll offer it again.
Any objection.
No.

It's admitted or readmitted.

BY MR. NELSON:

10

Q.

So let me have you explain the overall functionality

11

of Content Ads just using this diagram from PTX-222.

12

A.

13

is not a Google web page, it's a web page on a third-party

14

site.

15

Google to allow Google to display ads on it.

16

the Google ads, and I know it's really small.

17

user visits that site, a request goes to

Okay.

So what this shows here is a web page.

This

But that site has entered into an agreement with
So these are
So when the

18
19

Then it goes to the ad mixer.

The content ad

20

mixer is very similar to the ad mixer in Search Ads.

21

essentially it retrieves information about this page --

22

this page is about Java, the programming language, for

23

instance -- and retrieves a set of candidate ads.

24
25

And

Once it has a set of candidate ads, it conducts
a similar auction to determine which ads the user is most

722

Pazzani - direct
1

likely to click on and this auction involves a profile of

2

the user from Kansas as well as other information, and then

3

again the same maximization of revenue happens.

4

sent back through

The ads are

to that web page.

And all of that happens very quickly, within a

5
6

half second or so.

7

Q.

And what is the Kansas up there on slide 105?

8

A.

You probably remember there was a witness that drew

9

in during the deposition, so the Google document did not

10

contain Kansas but he updated the document to show that it

11

includes the user information from Kansas.

12

Q.

And let me have you turn to PTX-223 in your binder?

13

A.

Yes, I've got it.

14
15

It's a Google document entitled CUBAQ Overview;
where CUBAQ stands for Content User-Based Ads Quality.

16
17

MR. NELSON:

And let me offer PTX-223 into

evidence.

18

MR. VERHOEVEN:

No objection.

19

THE COURT:

20

(PTX-223 is admitted into evidence.)

21

MR. NELSON:

It's admitted.

Can you pull up slide 106, please?

22

BY MR. NELSON:

23

Q.

What does CUBAQ stand for, first of all?

24

A.

The Content User-Based Ads Quality.

25

Q.

What is shown on the pullout of this document?

723

Pazzani - direct
1

A.

On the pullout is actually shown the profiler.

2

the user profile is constructed from the ads, the sites the

3

user has visited.

4

section, it knows you learned about -- you like cars because

5

there is Google ads on the Los Angeles Times car section.

6

Also, the ads you have seen and the ads you have clicked on.

7

So

So if you go to the Los Angeles Times car

And then once you have that user profile, it

8

goes into the auction where the probability of the

9

click-through rate is multiplied by the cost per click to

10

determine what the winning ads are.

11

MR. NELSON:

Finally, let's talk for a minute

12

about YouTube.

13

Content Ads aspect of YouTube.

14

Can I have the next slide?

And really the

And may I have the next slide, please?

15

BY MR. NELSON:

16

Q.

17

did Mr. Nemeth say about YouTube?

18

A.

19

works, it's no different from the way it works on any other

20

site.

21

on YouTube but it uses the same mechanism in essence that it

22

uses for the Los Angeles Times or my blog.

23

Q.

And let me have you look in your notebook at PTX-226.

24

A.

Yes.

25

Q.

Can you identify that document?

And you may remember Mr. Nemeth from yesterday.

What

On YouTube, he said that the way Adwords, Content Ads

So Google owns YouTube and Google displays Google ads

724

Pazzani - direct
1

A.

PTX-226 is a letter to Mark Nelson from Google's

2

attorneys.

3

Q.

Thank you.

4
5

MR. NELSON:

I'd like to offer PTX-226 into

evidence with the clarification we discussed this morning.

6

MR. VERHOEVEN:

Your Honor, this is subject to

7

the morning's discussion and we have not seen a copy of the

8

document as we discussed to be provided.

9

this motion?

10
11

THE COURT:

Can we reserve on

You can reserve, but subject to

that, you're okay with the exam going forward?

12

MR. VERHOEVEN:

13

THE COURT:

14

MR. NELSON:

Yes, Your Honor.

Okay.
I'd like to put up PTX-226 but just

15

the portion that is here that relates to Mr. Nemeth.

16

BY MR. NELSON:

17

Q.

18

to Mr. Nemeth?

19

A.

20

are two methods for identifying and retrieves ads to display

21

on YouTube, using the DoubleClick system.

22

work, then it uses the AdSense For Content system.

23

further made it clear that if the DoubleClick system does

24

not provide adequate advertisements, then YouTube is treated

25

as any other publisher in the Adsense for Content system.

Can you tell me what Google's counsel said relating

Yes.

It said that Mr. Nemeth made clear that there

And if they don't
And he

725

Pazzani - direct
1

So there is a special advertising system for Google and

2

YouTube, and if that one doesn't work, then it uses AdSense

3

For Content and that is what we're alleging infringes.

4

MR. NELSON:

5

The next slide.

6

BY MR. NELSON:

7

Q.

8

about Content Ads --

9

A.

And let me put up PTX-110, please.

And Mr. Zamir testified yesterday.

What did he say

Well, he described in a little --

10
11

I'm sorry.

THE COURT:

Dr. Pazzani, you have to wait until

the question is done.

12

THE WITNESS:

Sorry.

13

BY MR. NELSON:

14

Q.

What did Mr. Zamir say about YouTube for Content Ads?

15

A.

He described in general about how YouTube for Content

16

Ads worked.

17
18

It's based on the

19

DoubleClick cookie, stores this information in Kansas, and

20

the

21

saw, serves their ads, and the infrastructure for the most

22

part is the same as the Content Ads.

23

Q.

24

infringes; is that right?

25

A.

that ad mixer that was on the diagram we

And you offered an opinion that YouTube Content Ads

Yes.

Pazzani - direct

726

1

Q.

And is that opinion based on the way the Content Ads

2

works?

3

A.

4

works the same as other content ad sites.

5

Q.

6

discuss in the next portion of your presentation, is your

7

opinion relating to the YouTube Content Ads going to be the

8

same as for Content Ads?

9

A.

Yes, it is.

10

Q.

I just don't want to ask the same questions all the

11

time relating to YouTube.

Yes.

So I relied on the Google witnesses who say it

For each of the claim elements that we're going to

12

May I have slide 111, please?

13

Can you just describe for now a summary of your

14

testimony from this morning as to what is depicted on this

15

slide?

16

A.

17

Again, the patent describes monitoring user interactions,

18

updating user specific data files, estimating parameters,

19

analyzing documents, estimating the probability the user is

20

interested in those documents, and providing personalized

21

services.

22

Q.

And what is depicted on the right?

23

A.

The right summarizes how the Google system operates.

24

It operates by transparently monitoring the user, updating

25

in Kansas information about the user.

Yes.

So this slide depicts Figure 2 from the patent.

727

Pazzani - direct
MR. NELSON:

1

Can I get you to blow up the boxes

2

here as he talks about each one, please?

3

BY THE WITNESS:

4

A.

Just the right side is fine.
So this shows that Google is transparently

5
6

monitoring the user.

It updates user specific data files,

7

the files in Kansas.

It estimates the parameters of the

8

learning machine.

9

profiles.

10

These are the weights associated with the

It analyzes documents.

It determines which

categories they belong into, whether it's

11

It uses a number of methods to estimate the

12

probability the user is interested in the document.

13

then it uses that probability to reorder the search results.

14

Q.

15

And

Thank you.
MR. NELSON:

With this, I'd like to turn to the

16

next portion of Dr. Pazzani's presentation now and add some

17

new notebooks.

18

THE COURT:

19

THE WITNESS:

20

THE COURT:

21

That's fine.
Can we take one away?

If he doesn't need all of those, you

might want to give him some space by removing some.

22

MR. NELSON:

23

(Binders passed forward and other binders taken

24
25

Yes.

off witness stand.)
MR. NELSON:

So may I have slide 1 of the next

728

Pazzani - direct
1

presentation?

2

BY MR. NELSON:

3

Q.

4

opinions.

5

more detail on an element-by-element basis of the claims.

6

And I want to start with the '040 patent, and I know a lot

7

of the '276 claim elements are very similar, and so we'll

8

work through that a little more quickly.

9

going to be some heavy going so we'll try to move through it

I'd like now to talk about, you have given your
I'd now like to talk about your opinions in a lot

10

as quickly as we can.

11

A.

And I know this is

Okay.

12

MR. NELSON:

So let's talk about claim 1 first

13

of the '040 patent.

14

please.

15

BY MR. NELSON:

16

Q.

So this is claim 1 of the '040 patent; right?

17

A.

That's correct.

18

Q.

And let's talk about the preamble first.

19

A.

Okay.

20

And particularly may I have slide 3,

Can you zoom in on that?
MR. NELSON:

Can I have the slide blown up?

21

you have a pullout?

22

BY MR. NELSON:

23

Q.

Can you read that?

24

A.

Sure.

25

computer-implemented method for providing automatic,

So this claim 1, the preamble is:

A

Do

Pazzani - direct

729

1

personalized information services to a user u, the method

2

comprising:

3

Q.

What is the highlighting?

4

A.

The highlighted terms are the terms the Court has

5

construed, provided a definition for.

6

Q.

7

patent, claim 1 from PTX-1; correct?

8

A.

9

And then it's the six steps below.

And just to make clear, so this is a copy of the

That's correct.
MR. NELSON:

And can you put up the next slide,

10

please?

11

BY MR. NELSON:

12

Q.

And what does this slide represent?

13

A.

The slide defines the word "user."

14
15

"A user is a person operating a computer or the
associated representation of the user."

16

MR. NELSON:

And may I have the next slide,

17

please.

18

BY MR. NELSON:

19

Q.

20

What does this document identify the user as?

21

A.

22

just a minute ago.

23

associated with signed-in users.

24

created accounts with passwords, and they then have a unique

25

ID.

And this is a portion of previously admitted PTX-576.

This describes a user.

We have seen the document

This is the Gaia ID, the identifier
Those are the users who

730

Pazzani - direct
1

Q.

Let me have you turn in your notebook to PTX-365,

2

please.

3

A.

(Witness complies.)

4

Q.

Did you find it?

5

A.

Yes, I did.

6

Q.

What is that document?

7

A.

It's the document "

8

previously been entered in.

9

Q.

Is this a Google document?

10

A.

Yes, it is.

11
12

MR. NELSON:

I'd like to move PTX-365 into

MR. VERHOEVEN:

Your Honor, I don't have an

objection but I think it's already been admitted.

15
16

in a Nutshell" that had

evidence.

13
14

Okay.

THE COURT:
readmitted.

Okay.

It's either admitted or

Thank you.

17

(PTX-365 is admitted into evidence.)

18

MR. NELSON:

Can you put that up on the board?

19

BY MR. NELSON:

20

Q.

21

respect to this document?

22

A.

23

Google places on the user's computer, and that cookie

24

contains a user identifier.

25

Q.

And what has Google identified the user as with

Well, this describes the

ID, the cookie that

And this, we're talking about Google Search here;

731

Pazzani - direct
1

right?

2

A.

Yes, Google Search and Google Search Ads use the
ID.

3

Again, it says it's used as a key for server

4

side data storage.

5

servers store data associated with that user by this

6

identifier.

7

What that means is that the Google

MR. NELSON:

Let me turn you in your notebook to

8

PTX-1312.

9

A.

(Witness complies.)

10

Q.

Can you identify that document?

11

A.

Yes.

12

Migrating Ads in Search to Use

13

Identifiers.

14

Q.

Is it a Google document?

15

A.

It's a Google document, yes.

16
17

Okay.

It's a new document, Work in Progress.

MR. NELSON:

UID.

That is User

I'd like to move PTX-1312 into

evidence.

18

MR. VERHOEVEN:

No objection, Your Honor.

19

THE COURT:

20

(PTX-1312 is admitted into evidence.)

21

MR. NELSON:

It's admitted.

Can you put up the next slide,

22

please.

23

BY MR. NELSON:

24

Q.

What does those document further tell you?

25

A.

This describes some user keys in Kansas.

These are

732

Pazzani - direct
1

the keys used to identify an enduser in Kansas.

2

of an ID type, such as the prefID, the

3

Gaia ID.
THE COURT:

4

It consists
ID, or the

Dr. Pazzani, let me remind you

5

again.

For the court reporter's benefit at least, please

6

wait until the question is complete before you begin to

7

answer.

Okay?
THE WITNESS:

8
9

Okay.

My wife is a court reporter

so she has taught me that.

10

BY MR. NELSON:

11

Q.

Let me turn you back to PTX-140, please.

12

A.

140.
Yes.

13
14

It's actually

an e-mail between several Google employees.

15
16

This is a Google document.

MR. NELSON:

And I'd like to offer PTX-140 into

evidence.

17

MR. VERHOEVEN:

No objection, Your Honor.

18

THE COURT:

19

(PTX-140 is admitted into evidence.)

20

MR. NELSON:

It's admitted.

Can you pull up slide 8, please.

21

BY MR. NELSON:

22

Q.

Who is Karthik Gopalratnam?

23

A.

He is one of the gentlemen, the Google engineers

24

interviewed yesterday.

25

Team.

He works in the Search Ads Quality

733

Pazzani - direct
1

Q.

What does this document indicate about users?

2

A.

That for signed-in users, they store data in the Gaia

3

tables.

4

Zwieback tables.

5

whether you have an account or whether you are not signed

6

into Google.

And for signed-out users, they store data in the
So, again, two different IDs depending on

MR. NELSON:

7
8

slide, slide 9.

9

And let me have you put up the next

And pull it out.

BY THE WITNESS:

10

A.

So we have seen this before.

11

cookie used by Google Search ads.

12

identifying the user.

13

Q.

And this was replaced by

14

A.

Yes.

15

Q.

Let me turn you in your notebook to PTX-407.

16

A.

Okay.

17

User-Based Quality.

19

It's another way of

It's defined as the prefID cookie.
correct?

It is a Google document entitled Content

MR. NELSON:

18

This is the prefID

I'd like to offer 407 into

evidence.

20

MR. VERHOEVEN:

No objection, Your Honor.

21

THE COURT:

22

(PTX-407 is admitted into evidence.)

It's admitted.

23

BY MR. NELSON:

24

Q.

And what does this document say?

25

A.

Well, it describes the DoubleClick cookie and how

Pazzani - direct

734

1

it is used in Content Ads.

2

DoubleClick cookie, they have access not only to the current

3

page being visited by the user but also all the browsing

4

history of the user.

5

Q.

And can you look in your notebook to PTX-406?

6

A.

Yes.

7

Q.

And what is that document?

8

A.

It is a test ad serving privacy policy and cookies,

9

cookie opt-out.

10
11

Now that Google is using this

It's a Google document.

MR. NELSON:

I'd like to offer Exhibit 406 into

evidence.

12

MR. VERHOEVEN:

No objection.

13

THE COURT:

14

(PTX-406 is admitted into evidence.)

15

MR. NELSON:

It's admitted.

Can I have the next slide, please?

16

BY MR. NELSON:

17

Q.

And what does the user ID there represent?

18

A.

The user ID here is one of those 64-bit numbers.

19

What it is showing is when an ad is served by the Google

20

server, Google records some information each time you view

21

an ad.

22

recording.

23

Q.

24

term "user" is used elsewhere in the claims; is that right?

25

A.

And it shows an example of the information its
The ad has an ID and the user has an ID.

Now, for all these different user identifiers, the

Yes.

It's used in almost every claim.

Pazzani - direct

735

1

Q.

And is it your opinion that these user identifiers

2

are the representations of the user throughout the claims?

3

A.

4

and this is an identifier that represents that user.

5

Q.

6

much time does it take for the results to come back?

7

A.

8

that's about my observation as well.

9

Q.

Yes, there is an enduser who is using the computer,

When you run a Google Search and get results, how

I think they try for less than half a second and
It might be a second.

Is there any human intervention during that period of

10

time?

11

A.

12

anything.

13

Q.

14

Search Ads as well?

15

A.

16

done automatically by Google servers.

17

Q.

And are all of these computer systems?

18

A.

Yes.

19

Q.

And do all of these accused systems provide

20

personalized information services?

21

A.

22

couple hours.

23

Q.

24

profiles that we discussed?

25

during the time they're created?

There's no time for any Google employee to do

Is that the same, is it true for Content Ads and

Yes.

Yes.

The results come back really quickly if it's

We've discussed personalization the first

What about with respect to the creation of the
Is there human intervention

736

Pazzani - direct
1

A.

No.

So some of the profiles,

,

2

are created on the fly, while the user is typing queries.

3

It's created between the time you type the query and it

4

returns the results.

5
6
7

Q.

8

please.

9

A.

10

And let me have you turn in your notebook to PTX-632,

Yes.

MR. NELSON:

I'd like to offer PTX-632 into

evidence.

13

MR. PERLSON:

14

THE COURT:

15

It's entitled

automatic profile generation and monitoring.

11
12

Of this is a Google document.

No objection.

It's admitted.

(PTX-632 was admitted into evidence.)

16

MR. NELSON:

Put the slide up.

Slide 13,

17

please.

18

BY MR. NELSON:

19

Q.

What does this document talk about?

20

A.

This document discusses the automatic profile

21

generation and it basically describes the profile generation

22

and monitors framework used in generating personalized

23

search results.

24

part here.

25

Q.

And, again, the automated is the important

That's one of the claim terms.

And let me put up slide 14.

And does this slide,

Pazzani - direct

737

1

what does this slide represent?

2

A.

3

of claim 1 that Google is a computer implemented method for

4

providing automated personalization.

5

Q.

6

each of the Google systems contain that element, perform

7

that element?

8

A.

9

checkmark for Search Ads, there's a search ad, and the

Well, it summarizes my opinion on just the first part

And do the checkmarks represent that you believe that

Yes.

The checkmark for search box there, the

10

checkmark for Content Ads and YouTube.

11

Q.

12

slide 15, and the next element, the transparently monitoring

13

the user interaction with data element.

All right.

14

Let's turn to the next, let me turn to

May I have slide 16, please, and the pullout.

15

And so this is Element A now, transparently monitoring.

16

you explain this element?

17

A.

18

interactions with data while the user is engaged in normal

19

use of a computer.

20

transparently means something like without any extra user

21

effort and user is defined as it was before.

22

Q.

And what is the normal use of a computer?

23

A.

Well, things like browsing and searching and e-mail

24

are the normal uses of the computer.

25

Q.

Yes.

Can

This is transparently monitoring user

So this is claim 1(a) where

And does Google Search, Search Ads and Content Ads,

738

Pazzani - direct
1

each do their transparently monitoring while the user is

2

engaged in normal use of a computer?

3

MR. VERHOEVEN:

4

THE COURT:

Objection, Your Honor.

Leading.

Overruled.

5

BY MR. NELSON:

6

Q.

Go ahead and answer.

7

A.

Yes, they do.

8

Q.

Let me put up -- let me direct you in your notebook

9

to PTX-11.

10

A.

Yes.

11

personalized search."

12
13

It's a Google document entitled "Cookie-Based

MR. NELSON:

I'd like to move PTX-11 into

evidence.

14

MR. VERHOEVEN:

15

THE COURT:

No objection, Your Honor.

It's admitted.

16

(PTX-11 was admitted into evidence.)

17

MR. NELSON:

Can you pull up the pullup?

18

BY MR. NELSON:

19

Q.

And what does this document tell you?

20

A.

It goes into more detail about what is transparently

21

monitored.

22
23
24
25

These are the
that are recorded as part of a normal web search

and they're associated with these cookie-based identifiers,
, Gaia, and retained for a finite period of time,
.

739

Pazzani - direct
1

Q.

And let me have slide 18, please.

2

Mr. Horling from yesterday.

3

A.

What did he say?

So the two-line quote from that video is:
"So the system watches what the user is

4
5

This is

doing and tracks it?"

6

And his answer is:

7

Q.

8

did he say about the Search Ad system?

9

A.

10

Let me have slide 19.

"At a high level, yes."

This is Mr. Gopalratnam.

What

The Search Ad system he talked a little bit about and

he confirms that they monitor the

11
12
13
14

Q.

Let's go back to search for a minute.

15

have to do anything for this monitoring to happen?

16

A.

No.

17

Q.

Does the user -- they don't have to sign in?

18

don't have to do anything?

19

A.

That's correct.

20

Q.

What about for Search Ads?

21

even going on?

22

A.

It's effortless to the user.

23

Q.

What about for content ads?

24

user there as well?

25

A.

Yes.

Does the user

It's effortless to the user.
They

Is the user aware this is

Is it transparent to the

The user doesn't have to do any steps to get

740

Pazzani - direct
1

his information, his or her information recorded in the

2

Google databases.

3

Q.

4

Actually, just go to the next slide.

5

been admitted.

And let me direct your attention to PTX-395.
This one has already

6

This is a slide about Search Ads.

7

this slide say with respect to what's monitored in Search

8

Ads?

9

A.

10

What does

Well, this describes again what is monitored in

Search Ads,

11
12
13

Q.

And what is UBAQ on the bottom?

14

A.

UBAQ is actually a profile of the user in Search Ads,

15

and that is a summary of their activity in the

16

It expresses what the user is interested in and what the

17

user is not interested in.

18

Q.

19

Mr. Ponnekanti from yesterday.

20

monitored in Search Ads?

21

A.

This is Content Ads, I believe.

22

Q.

I'm sorry.

23

A.

And he said that Content Ads monitors

24
25

And let me put up slide 21, please.

This is

What did he say what was

Content Ads.

741

Pazzani - direct
1
2
3
4

Q.

And let me put up slide number 22.

5

from yesterday.

6

A.

7

Ads looks at

Yes.

This is Mr. Zamir

Can you summarize what he said?

The quick summary is he said that Content

8
9
10

Q.

And just one more.

11

PTX-404.

12

A.

13

where AFC stands for Ad Sense per content.

Okay.

14
15

Let me direct your attention to

It's a Google document AFC user profiler,

MR. NELSON:

I'd like to offer Exhibit 404 into

evidence.

16

MR. VERHOEVEN:

No objection, Your Honor.

17

THE COURT:

18

(PTX-404 was admitted into evidence.)

It is admitted.

19

BY MR. NELSON:

20

Q.

21

Exhibit 404 say about what content --

22

A.

23

says, we, that is Google, looks at the following activities

24

for each user:

25

Q.

Can you put that on the screen?

And what does

It goes into more details about Content Ads.

It

And let me put, direct your attention to PTX-370.

742

Pazzani - direct
1

And what is Exhibit 370?

2

A.

3

read, write web.

370 is a document from a website, I believe, called

4
5

Its author is Greg Linden.

MR. NELSON:

And I want to offer Exhibit 370

into evidence.

6

MR. VERHOEVEN:

7

THE COURT:

No objection, Your Honor.

Admitted.

8

(PTX-370 was admitted into evidence.)

9

BY MR. NELSON:

10

Q.

Who is Greg Linden?

11

A.

Greg Linden was a former Amazon employee.

12

person that wrote Amazon's personalization system.

13

Q.

14

topic?

15

A.

16

explicit to use it.

17

personalized search.

18

the transparently monitoring.

19

Q.

20

your opinion whether Google's search, Search Ads and Content

21

Ads and YouTube Content Ads practice all of the elements of

22

claim 1 of '040 claim, element 1(a)?

23

A.

24

searches, your result clicks, the ads that you've seen, the

25

ads that you've clicked on, whether it's Search Ads or

He is the

And what does Mr. Linden say about Google on this

He says that searchers don't have to do anything
It here is referring to Google
It's all implicit.

And let me turn to slide 25.

Yes.

So he's getting at

And can you summarize

So Google keeps track of things like your

743

Pazzani - direct
1

Content Ads, anywhere including YouTube.

2

Q.

Is it your opinion that that element is met?

3

A.

Yes.

4

(Pause while counsel conferred.)

5

MR. VERHOEVEN:

6

THE COURT:

7

I apologize, Your Honor.

You all can meet and confer.

That's

fine.

8

(Pause.)

9

MR. VERHOEVEN:

Your Honor, I don't want to

10

disrupt the proceedings, but I thought we had agreed that

11

the left-hand column would be modified to have all the words

12

and it was represented it would.

13

soon as possible, I won't interrupt the proceedings.

14
15

THE COURT:

If we could fix that as

Will you agree you'll fix that when

you have a chance?

16

MR. NELSON:

Yes, we are.

There are some

17

larger boards that we have later that have them all.

18

clear we're going through all of the elements of the claims

19

here.

20

THE COURT:

It's

We don't need any more argument

21

about it, but we'll fix it, and when we do, we'll let the

22

ladies and gentlemen of the jury know that we've replaced

23

it.

24
25

MR. VERHOEVEN:
BY MR. NELSON:

Thank you, Your Honor.

Pazzani - direct

744

1

Q.

So to summarize your opinion on claim element 1(a) --

2

A.

Yes.

3

Search -THE COURT:

Dr. Pazzani, I know you had to wait

4

through all of that, but it's important that you wait until

5

the question is asked before you answer it.

6

Go ahead.

7

BY MR. NELSON:

8

Q.

9

your opinion with respect to whether Google's search Search

Let me re-ask it.

Dr. Pazzani, can you summarize

10

Ads, Content Ads and YouTube practice each element of the

11

'040, claim limit 1(a)?

12

'040 patent?

13

A.

Yes, they do.

14

Q.

Let's turn to element B.

Each aspect of element A of the

15

MR. NELSON:

16

you pull out Element B.

17

BY MR. NELSON:

18

Q.

And can you read Element B to the jury?

19

A.

Yes.

20

Updating user-specific data files, wherein the user-specific

21

data files comprise the monitored user interactions with the

22

data and a set of documents associated with the user.

23

Can I get slide 27, please.

Can

Element B is the second element of claim 1.

MR. NELSON:

24

BY MR. NELSON:

25

Q.

And may I have slide 28, please.

And can you read what is on slide 28 and tell me what

Pazzani - direct

745

1

it is?

2

A.

3

terms.

4

interactions with data and a set of documents associated

5

with the user.

Yes.

So this is the Court's construction of two

User-specific data files is the monitored

6

And the Court has defined monitored user

7

interactions with the data to be, the collected information

8

about the user's interactions with the data.

9

Q.

10

And let's talk first about that aspect, the monitor

user interactions with the data.

11

MR. NELSON:

Could I get slide 29.

12

BY MR. NELSON:

13

Q.

14

PTX-15 that we talked about earlier.

15

this document is talking about?

16

A.

17

is updated and it's -- Kansas contains the Gaia ID, the ID

18

of the signed-in user, and it updates the data that it

19

stores, or the web queries as well as the results clicks.

20

Q.

And what about for the PrefID?

21

A.

That same data is stored for the PrefID tables, the

22

earlier once, and now the

23

Q.

24
25

And this is a pullout.

Yes.

And this is a portion of
Can you describe what

So Kansas stores the user-specific data and it

as well.

And let me have slide 30, please.
And so this is -- first of all, can you

identify the data on the left part, that's the PTX-373?

746

Pazzani - direct
1

A.

Yes.

2

think this is

3

This is from one of the

interactions.

.

.

I

And this shows the

So the query migration

4
5
6

Q.

There's also the

7

next, I think.
And what about the portion up on the top

8
9

that we will discuss

there?

What is that intending to represent?

A.

This is representing the Kansas database.

17

Q.

Let me turn to the next slide, please.

18

talk now about the set of documents associated with the

19

user.

Can you tell the jury how that was defined?

20

A.

Yes.

21

a group recollection of documents associated with the user.

22

Q.

23

the document?

24

A.

25

any type of media.

10
11
12
13
14
15
16

And so let's

A set of documents associated with the user is

And what is a document?

What is the second part of

A document is an electronic file, including text, or

747

Pazzani - direct
MR. NELSON:

1

And let me have slide 32, please.

2

BY MR. NELSON:

3

Q.

4

explain what a document ID is?

5

A.

6

to uniquely represent the document.

7

Google finds on the web and indexes has this unique

8

document.

9

sometimes we'll associate documents with the user by

And this is PTX-372 that we saw earlier.

Yes.

Can you

A document ID is an identifier that Google has
So each document that

And it's generated in part from the URL, so

10

associating the URL with the user ID.

11

the document ID associated with the user ID.

12

Q.

13

the left portion here the Q and the RC information that is

14

part of PTX-373, what that data is?

15

A.

And let me turn to slide 33.

Yes.

Sometimes it will be

And can you identify on

So, again, this is data from the

16
17
18

It's an address of

19

the document that is associated with the user ID.

20

You store

it in the database by

21
22

Q.

And in Google Search, what is the document?

23

A.

The document is the document on the website, like the

24

Wikipedia website, and the identifier is stored in Google's

25

databases.

748

Pazzani - direct
MR. NELSON:

1

And let me have the next slide,

2

please.

3

BY MR. NELSON:

4

Q.

5

to show?

6

A.

7

with the document by means of the document ID.

8

is something out there on the web.

9

identifier for that document.

And can you just explain what this slide is intended

Yes.

This slide depicts how you associate the user
The document

Google's database has an

It stores that identifier

10

with the user identifier to associate the user with that

11

document.

12

Q.

And so what is the user ID a represent here?

13

A.

Well, that would be this

14

The document ID would be one of those numbers or the

15

address, and then the document would be the document out on

16

the web.

17

Q.

18

associated with the user?

19

A.

20

documents, for example, in the

21

web results column that shows the web impressions.

22

another association.

23

see them in a minute.

24

Q.

25

right?

or the Gaia ID.

And how does Google Systems create a set of documents

It stores associated with the user ID a set of
There's also a

There are some others as well.

That's
You'll

So the URLs themselves aren't the documents; is that

Pazzani - direct
No.

749

1

A.

The URLs are the address of the document, but,

2

again, Google has analyzed that document, understands the

3

content, can retrieve that address by the content of the

4

document, and then using that content, construct a search

5

results page so the user could click on that URL and get to

6

the document.

7

Q.

8

next slide, please.

9

Let me have the pull out.

And let me turn your attention to -- let me have the

10

So this is a portion, then, of PTX-375.

Can you explain what's going on in this slide?

11

A.

Yes.

This is another part that we have not discussed

12

yet of Kansas.

Kansas has a

13
14
15
16
17
18
19
20
21

Q.

And just for the record, the pullout is PGLPUM11643.

22

A.

So these documents are also associated with that ID

23

and in that big binder.

24

Q.

25

GGLPUM114940.

This is another portion of the same document,
Can you explain what's going on here?

750

Pazzani - direct
1

A.

Yes.

2
3
4
, has the dates that this was

5
6

done, on July 1st of 2010.

7

Q.

8

from yesterday.

9

A.

And let me turn to slide 37.

And this is Mr. Horling

What is Mr. Horling saying?

Well, the quote is, does Kansas -- the way the system

10

currently works, does Kansas use two types of user

11

identifiers; one is a Gaia ID and one is a

12
13

?

And he confirmed, yes, those two user
identifiers are used in Kansas.

14

MR. NELSON:

And let's turn to Search Ads next.

15

Can I get slide 39, please?

16

BY MR. NELSON:

17

Q.

18

What is this document tell you about the updating user

19

specific data files?

20

A.

21

data and associates it with the user ID.

22

the queries, the ad clicks, result clicks, and ad

23

impressions that get stored.

24
25

This is a document we have seen before, PTX-395.

This is where Search Ads stores the user specific

MR. NELSON:
BY MR. NELSON:

And, again, it's

And let me turn to slide 40.

751

Pazzani - direct
1

Q.

And what did Mr. Gopalratnam say about this?

2

A.

He is describing updating the Kansas database:
How often does Kansas update the information

3
4

associated with the user ID?

5
6
And then you asked:

7

Each time?

8
9
MR. NELSON:

10
11

BY MR. NELSON:

13

Q.

14

on in this slide?

15

A.

16

Let's go to

slide 41.

12

And let's go back.

only been looking at the

And so this is again part of PTX-375.

So this is the

What is going

So far, we've

17
18
19
20
21
22
23
24
25

MR. NELSON:
please.

Let me turn to the next slide,

752

Pazzani - direct
1

BY MR. NELSON:

2

Q.

3

you tell the jury that again?

4

A.

5

text or any type of media."

So this is the Court's definition of "document."

Yes.

"Document" is "an electronic file including

6

MR. VERHOEVEN:

7

THE COURT:

8

MR. VERHOEVEN:

9

two slides.

Can

Your Honor?

Yes.
We have an objection to the next

And we've talked about it briefly.

I don't

10

think it has been resolved; and I respectfully request a

11

short sidebar.

12

THE COURT:

13

(Sidebar conference held.)

14

THE COURT:

15
16

All right we'll have a sidebar.

All right.

So what is the

objection?
MR. VERHOEVEN:

I apologize, Your Honor.

We

17

alluded to this previously but they have been going along

18

with this fine.

19

construction with more claim construction interpretation,

20

the terminology "document," and then this is the one you

21

have seen before with what a "file" is, Your Honor.

22

But now they're going to interpret your

We object.

This is essentially in our view, you

23

issued a claim construction ruling on "document" and now

24

they're construing the claim construction and you used the

25

word "file" and now they're trying to construe the word

753

Pazzani - direct
1

"file" within it and we would submit it is inappropriate for

2

an expert witness to tell the jury what these meanings are.

3

And, in fact, we think that there is no need for

4

this.

The witness has already said he thinks the document

5

element is met.

6

then talk about the meaning of the words of your

7

construction to support that.

8

wants or alternatively, Your Honor, if you don't want to

9

proceed that way, then we think that it's a matter for the

And it's not appropriate for him to go and

He can say factually what he

10

Court to decide what a file is, not two experts fighting

11

over it.

12

THE COURT:

13

MR. NELSON:

What is your position?
Well, first of all, our position is

14

this objection has been waived.

15

Google last night as to what remained objectionable on Dr.

16

Pazzani's slide and this wasn't it.

17

the correspondence was limited to what Mr. Perlson presented

18

this morning:

19

We had correspondence with

It was very clear that

The two letters and also --

THE COURT:

We did talk about this before, but I

20

don't think I ruled.

21

did something to waive their objection?

22

You're saying subsequent to this, they

MR. NELSON:

Yes, Your Honor.

We sent an e-mail

23

last night I think in a summary of what was left in Dr.

24

Pazzani's slides that were objectionable?

25

back was the two letters and the deposition exhibit.

And what was sent
That

754

Pazzani - direct
1

was it.

2
3

THE COURT:

Assuming it's not waived, what is

your position?

4

MR. NELSON:

Our position is that this is

5

clearly in the expert report.

They've had their own expert

6

had an opportunity to opine back on it.

They had a chance

7

to take Dr. Pazzani's deposition on it.

They can cross him

8

on it.

9

testimony isn't necessarily the -- it is fairly low on the

And albeit on the claim construction, law expert

10

overall hierarchy.

11

to give.

12

This is perfectly fine testimony for him

If they want to cross, depending if this is in

13

Dr. Fox's report on it or not, they can put something on

14

assuming it's in his report, but that they've been on notice

15

since the time of Dr. Pazzani's report that there was a

16

dispute, there was a disagreement as to what the word "file"

17

meant.

18
19
20

THE COURT:

Do you think I need to construe it

then?
MR. NELSON:

I don't know if you need to

21

construe it or not, Your Honor.

22

the plain and ordinary meaning provides aspects of that.

23

And if they wanted this construed, they could have asked for

24

it to be construed later.

25

THE COURT:

I think that it's a word,

Is there anything further?

755

Pazzani - direct
1

MR. VERHOEVEN:

Yes, we did ask for a document

2

to be construed and we got a construction.

3

with the construction.

4

construction, Your Honor.

5

my opinion to violate O2Micro and its progeny.

6

They're unhappy

Now they're construing the
This is an invitation by them in

And as to the point that expert testimony has

7

not a lot of relevance, a little bit of relevance to claim

8

construction, that misses the point that I was making which

9

is that is for the Court, it's not for the jury.

10
11
12
13

Claim

construction is for the Court.
THE COURT:

Here is what we're going to do.

I'm

going to overrule the objection.
First off, on waiver, it has not been proven to

14

me that the objection was waived.

So I'm going to reach the

15

merits of the objection but I'm going to overrule and permit

16

the testimony from the plaintiff's expert provided, and I

17

assume this will be the case, that he will say that he is

18

applying the Court's claim construction on those terms that

19

we did construe.

20

I will give similar leeway to defendants on

21

cross as well as to present testimony from their expert

22

consistent with our claim construction.

23

right, if need be, to construe any term in dispute and to

24

instruct the jury accordingly, which, of course, may be

25

inconsistent in the end with what any expert may have

And I reserve the

Pazzani - direct

756

1

testified to as being his or her understanding of "file."

2

But that is the ruling.

3

Any questions about that?

4

MR. NELSON:

5

MR. VERHOEVEN:

6

THE COURT:

7

(Sidebar conference ends.)

8

THE COURT:

9

MR. NELSON:

No, Your Honor.
No, Your Honor.

Okay.

You can continue when you are ready.
Thank you, Your Honor.

10

BY MR. NELSON:

11

Q.

12

again -- or, no.

13

Court's construction to the jury?

14

A.

15

electronic file including text or any type of media."

16

Q.

And let me have you look at PTX-357 in your notebook.

17

A.

357 is a definition from the Random House Webster's

18

Unabridged Dictionary.

So, Dr. Pazzani, can you read the jury's construction

Yes.

19
20
21
22

Let me start over.

Can you read the

The Court has defined the "document" to be "an

MR. NELSON:

And I want to move 357 into

evidence.
MR. VERHOEVEN:

Subject to the sidebar, Your

Honor, no objection.

23

THE COURT:

It's admitted.

24

(PTX-357 is admitted into evidence.)

25

MR. NELSON:

Can you put up a slide 43, please?

Pazzani - direct
1

757

Can you put up the slide of the Random House

2

Dictionary?

3

BY MR. NELSON:

4

Q.

And can you tell what is this slide?

5

A.

This slide is a definition of the word "file" from

6

the Random House Dictionary.

7

It says:

A file is a folder, a cabinet or other

8

container in which papers, letters, et cetera, are arranged

9

in convenient order for storage or reference.

10

Then it goes on to give the computer definition

11

of that:

A collection of related data or program records on

12

some input, output, auxiliary storage medium.

13

Q.

14

please.

Let me have you look in your notebook to PTX-1113,

15

MR. VERHOEVEN:

16

MR. NELSON:

17

THE WITNESS:

What was the number?

1113.
1113 is the Users-Based Ad

18

Quality.

19

BY MR. NELSON:

20

Q.

Is it a Google document?

21

A.

Yes, it is a Google document.

22
23
24
25

MR. NELSON:

I want to offer Exhibit 1113 into

evidence.
MR. VERHOEVEN:

No, I have no objection, Your

Honor, except my copy is illegible so if we can get a

Pazzani - direct
1

legible copy?

2

THE COURT:

3

MR. NELSON:

4

I'm sure one can be provided.
Yes.

I'm not sure the one we have

is a lot better, which you will see in a minute.

5
6

758

THE COURT:

All right.

Well, it's admitted and

you will give him the best copy you have.

7

(PTX-1113 is admitted into evidence.)

8

MR. NELSON:

9

And can you blow up the portion on the left,

Can I have slide 46, please?

10

please?

11

BY MR. NELSON:

12

Q.

All right.

13

A.

I think it's the animation that has it.

14

Q.

Yes.

15

the portion here that we have retyped.

16

A.

What we're going to be talking about here is

Okay.

17

I can read it, sort of.
MR. NELSON:

So can I have the blowup portion of

18

the slide now?

19

BY MR. NELSON:

20

Q.

And can you tell me what this document says?

21

A.

Yes.

22

Google ad has several elements.

23

text you see on the top of the ad.

24

also called the creative.

25

describe the ad.

This document describes Google ads, and a
It's a headline, that's the
The lines of text, it's

That's the few lines of text that

A display URL, that's the website that it

759

Pazzani - direct
1

is from, like Amazon or eBay.

2

that's the place that you will go to if you click on the ad.

3

Q.

Let me have you turn to PTX-399, please.

4

A.

(Witness complies.)

5

SmartAds for Smarties.

6
7

MR. NELSON:

And the destination URL,

399 is a Google document,

I'd like to offer Exhibit 399 into

evidence.

8

MR. VERHOEVEN:

9

THE COURT:

No objection, Your Honor.

It's admitted.

10

(PTX-339 is admitted into evidence.)

11

MR. NELSON:

Can you put Exhibit 399 up, please?

12

BY MR. NELSON:

13

Q.

14

right box?

15

A.

16

Find a Job:

17

keyword is what is called the ad text.

18

And so what is the Find a Job portion up in the upper

The Find a Job is the headline.

And then the text,

Search job listing by location, industry or

The www.hotjobs.com is the visible URL.

And

19

there, the components of the ads, so ads have identifiers.

20

You can think of them as being stored in a file cabinet and

21

associated with that ad identifier are these components, the

22

visible URL, the ad text, the headline, and also the

23

destination URL that is now depicted here.

24

Q.

And this is an electronic system; is that right?

25

A.

That's correct.

760

Pazzani - direct
1

Q.

And so these ads, is there something that glues them

2

together?

3

A.

4

another ad.

5

Q.

6

according to the Court's definition?

7

A.

Yes, they are.

8

Q.

And let's talk a little bit more about that.

9

me turn your attention to PTX-356.

Yes, the ad identifier is how you tell one ad from
It's how Google tells one ad from another ad.

And so, in your opinion, are ads electronic files

10

A.

11

So let

Fourth Edition.

12
13

MR. NELSON:

MR. VERHOEVEN:
issue, Your Honor.

16
17

I'd like to ask that PTX-356 be

moved into evidence.

14
15

356 is from Microsoft's Computer Dictionary, the

This is subject to the same

Subject to that, no objection.

THE COURT:
rights, it is admitted.

Okay.

Given that reservation of

You may proceed.

18

(PTX-356 is admitted into evidence.)

19

MR. NELSON:

Can we have slide 48 please?

20

BY THE WITNESS:

21

A.

22

complete named collection of information, such as a program,

23

a set of data used by a program, or a user-created document.

24

A file is the basic unit of storage that enables a computer

25

to distinguish one set of information from another.

48 is Microsoft's definition of a "file."

It's a

Pazzani - direct

761

1

Q.

And in your view, does an ad meet that -- meet the

2

Court's definition?

3

A.

4

identifier, and that ad identifier has ads associated with

5

it and only it.

Yes.

So there is a unit of storage, the ad

6
7

MR. NELSON:

Let me go back for a second.

I

8

skipped a portion here that I wanted to, I wanted to talk

9

about first.

10

Can I get slide 42 back up again?

11

BY MR. NELSON:

12

Q.

13

correct?

14

A.

Correct.

15

Q.

And so may I have you look in your folders to

16

PTX-401, please?

17

A.

Yes.

18

Q.

And what is PTX-401?

19

A.

PTX-401 is a Google document.

20

Roller Large Auction Ad Targeting.

So that is the Court's definition of "document,"

21
22

MR. NELSON:

It's called High

I want to move Exhibit 401 into

evidence.

23

MR. VERHOEVEN:

No objection.

24

THE COURT:

25

(PTX-401 is admitted into evidence.)

It's admitted.

762

Pazzani - direct
1

MR. NELSON:

Can I get slide 43, please.

2

BY MR. NELSON:

3

Q.

4

is Google -- well, what is this document say about

5

documents?

6

A.

So what is -- this is terminology of PTX-401.

Well, this document defines --

7
8

What

MR. VERHOEVEN:
objection.

For the record, Your Honor, same

Reservation to that.

9

THE COURT:

Fine.

Go ahead.

You can answer.

10

BY THE WITNESS:

11

A.

12

that is used by the SmartAds system.

13

that judges the probability the user would be interested in

14

the document.

Okay.

15

So this is a Google definition of "document"

And it says:

That is the system

A document is the content

16

associated with a single ad.

17

it's creative, it's landing page.

18

of the ad.

19

For instance, its customer ID,
The creative is the text

And we haven't talked about it yesterday but the

20

customer ID is the Google customer, the advertiser who pays

21

Google to put that ad there.

22

to associate that with the ad as well so Google can charge

23

that customer.

24
25

MR. VERHOEVEN:

And, of course, you would want

Objection to that testimony as

inconsistent with the Court's claim construction.

763

Pazzani - direct
1

THE COURT:

2

Okay.

MR. NELSON:

Disagree.

3

BY MR. NELSON:

4

Q.

5

system?

6

A.

7

associated with a single ad.

What does Google say a document is in the Search Ad

So let me read it again.

8
9

THE COURT:

Document:

The content

The objection is overruled.

BY MR. NELSON:

10

Q.

And in your opinion, is an ad a document according to

11

the Court's construction?

12

A.

13

ad.

14

Q.

15

different position that an ad isn't a document.

16

your understanding as well?

17

A.

18

in the opening argument.

19

Q.

Do you agree with it?

20

A.

No, I don't.

21

Q.

Let's just say we give them the benefit of the doubt.

22

Do you have an opinion as to whether ads are equivalent to

23

documents?

24

A.

Yes, I do.

25

Q.

And what is that you opinion?

Yes.

An ad is the content associated with a single

A document is the content associated with a single ad.
And it's my understanding that Google takes a

I have heard that argument, yes.

Is that

I think we saw it

It's pretty clear here.

764

Pazzani - direct
1

A.

Yes, they are equivalent.

2

Q.

Why is that?

3

A.

Can I show the slide?

4

Q.

Yes.

5

Let me put up slide 49.
And go ahead and give your opinion.

6

A.

So there is a test for equivalence.

And ads are not

7

substantially different from documents.

8

to documents.

9

electronic files in substantially the same way -- I'm sorry.

They're equivalent

They function substantially the same way as

10

They function substantially as electronic files in

11

substantially the same way to achieve substantially the same

12

result.

13

So, again, ads are substantially the same as

14

documents.

15

we saw that Google has databases that store ads.

16

databases that store web documents, the contents of web

17

documents.

18

They are indexed and stored electronically.

Google has

Ads are identified by their identifiers.

19

identifies the properties of ads.

20

So

Google

associated with ads.

21

They'll be categories

Google estimates a user's interest in them.

And

22

Google retrieves them, ranks them, and provides them to the

23

user.

24

Q.

25

the Content Ad system?

And is that opinion the same with respect to ads in

765

Pazzani - direct
1

A.

2

the same ads.

3

but there is not really a different way of.

4

advertisers let both of them happen.

5

Q.

6

element in the '276 patent?

7

A.

Yes, it is, where a document is used in '276 as well.

8

Q.

So let me have you turn to PTX-220.

9

A.

Okay.

10

Yes, both Content Ads and Search Ads really use about

Usually,

Is that opinion the same for this related claim

It's a Google document entitled Content Ads

Also Known As Adsense.

11
12

Advertisers can specify one system or another

MR. NELSON:

I want to move for PTX-220 to be

admitted.

13

MR. VERHOEVEN:

No objection.

14

THE COURT:

15

(PTX-220 is admitted into evidence.)

16

MR. NELSON:

It's admitted.

Can you put up the next slide.

17

BY MR. NELSON:

18

Q.

19

treated in Google systems?

20

A.

21

earlier we talked a little bit about how

22

document IDs for documents on the web.

23
24
25

What does those document tell you about how ads are

It talks about the

So
has the
Well,

766

Pazzani - direct
1
2

Q.

And let's talk, let's go to the next slide, please.

3

And talk about the Content Ad system briefly.

4

you look at PTX-403, please.

5

is PTX-403?

6

A.

7

second set of requests for admission.

And what is Exhibit 403?

What

Oh, PTX-403 is Google's response to the plaintiff's

8
9

May I have

MR. NELSON:

And I want to move PTX-403 into

evidence.

10

MR. VERHOEVEN:

11

THE COURT:

No objection, Your Honor.

It's admit.

12

(PTX-403 was admitted into evidence.)

13

BY MR. NELSON:

14

Q.

15

PUM to questions that PUM asked?

16

A.

Yes, they are.

17

Q.

And let me have you focus on response to request for

18

admission 6.

And so are these answers that Google gave us to, gave

19

What does Google say about the Content Ads

20

element?

21

A.

22

Google stores data associated with a DoubleClick cookie

23

regarding

24
25

Well, for a CUBAQ, the Content Ad Backlogging System,

767

Pazzani - direct

So when you have a web page like the Los

1
2

Angeles Times that displays Content Ads from the Los Angeles

3

Times site, it asks the Google site to put ads in there and

4

that's called the ad request.

5

Q.

6

notebook.

7

A.

Okay.

8

Q.

And what is that document?

9

A.

It is a document entitled overview, which describes

10

And let me direct your attention to PTX-408 in your

the Google AdSense System.
MR. NELSON:

11
12

I want to move PTX-408 into

evidence.

13

MR. VERHOEVEN:

14

THE COURT:

No objection.

It's admitted.

15

(PTX-408 was admitted into evidence.)

16

BY MR. NELSON:

17

Q.

Slide 403.

18

A.

That talks about part of the Content Ads, part of the

19

Kansas infrastructure and what it stores.

20

it stores clicks,

What does this document say?

And it says that

21

And they

22

are stored in the Kansas infrastructure.

That's updating

23

the user-specific files associated in this case with the

24

DoubleClick cookie.

25

Q.

And let me put up the next slide, please.

And this

768

Pazzani - direct
1

is Mr. Zamir from yesterday.

Can you just summarize what he

2

said?

3

A.

Yes.

6

Q.

And let me have PTX-401 again.

7

document we had up earlier.

He confirmed that the Content Ads updates

4
5

What does Google say is document is in the

8
9
10

And so this is the

Content Ad system?
A.

A document --

11

MR. VERHOEVEN:

12

THE COURT:

13

THE WITNESS:

Same objection, Your Honor.

The objection is noted.
A document again is the content

14

associated with a single ad, and so you would identify that

15

content by an ad ID -

16

Q.

Can you turn in your notebook to PTX-409, please?

17

A.

Yes.

18

Q.

And --

19

A.

It is a Google document entitled audience/interest-

20

based advertising, subtitled Kansas use in the

21
22
23

Q.

And can you put -MR. NELSON:

I'd like to move that Exhibit 409

24

be admitted into evidence.

25

MR. VERHOEVEN:

No objection.

769

Pazzani - direct
THE COURT:

1

It's admitted.

2

(PTX-409 was admitted into evidence.)

3

BY MR. NELSON:

4

Q.

And so what does this document indicate?

5

A.

It describes how the updates at Kansas are done and

6
7
8

Basically helps write to the Kansas database.

9

Q.

And for search, Search Ads and Content Ads as well as

10

YouTube Content Ads, are the user-specific data files

11

updated?

12

A.

13

for instance, in all cases.

14

Q.

15

tell me, can you summarize your opinions with respect to

16

whether Google's search, Search Ads, Content Ads and YouTube

17

content ads practice claim element 1(b) of the '040 patent?

18

A.

19

and updates them as the user uses the Google System.

20

Q.

21
22
23
24
25

Yes.

The Kansas data files contain the

And let me turn to slide 58, please.

Yes.

And can you

So each of them stores information in Kansas

Thank you.
Now getting ready to turn onto Element C.

I'm not sure when the lunch break is.
THE COURT:

Do you think you need more than

15 minutes for 1(c)?
MR. NELSON:

Yes.

Pazzani - direct
1

THE COURT:

2

break now.

3

All right.

770

Then let's take a lunch

gentlemen of the jury.

4
5

I know lunch is here for the ladies and

No talking about the case during the break and
we'll bring you back in about a half-hour.

6

(The jury was excused for a luncheon recess.)

7

THE COURT:

Before we break, when we come back

8

from lunch, is it possible, I have a couple questions about

9

the Twersky deposition designation objection, so if you can

10

have whoever is able to speak to that here, we may have a

11

few questions for you.

12

But we'll take a recess.

(Luncheon recess taken.)

13

-

-

-

14

Afternoon Session - 12:53 p.m.

15

THE COURT:

Before we bring the jury in, just

16

one quick question on Twersky.

17

451, I guess it's Google's argument that the

18

counter-designations proposed by PUM is untimely and

19

improper.

20

So first from Google.

21

At the designation at page

I just wanted to briefly understand that better.

MR. PERLSON:

Well, I actually -- I know most of

22

these, but that was the one that I wasn't involved in.

23

know all the substantive stuff.

24

THE COURT:

25

MR. PERLSON:

I

Right.
I think that there was -- it was

771

Pazzani - direct
1

just too late.

2

THE COURT:

The argument, I guess, is, it wasn't

3

timely, consistent with the whole procedure for disclosing

4

to one another?

5
6

MR. PERLSON:

THE COURT:

I

All right.

Let me see what PUM has

to say.

9
10

That's my understanding.

don't know why it's timely.

7
8

Yes.

MS. MURPHY:

Good afternoon, Your Honor.

Regina

Murphy for PUM.

11

So, yes.

In terms of our exchange procedure, it

12

was untimely after the meet and confer to try to address

13

what we understood were Google's objections.

14

withdraw our objections, we proposed we could

15

counter-designate that portion.

16

time to disclose originally.

17
18

THE COURT:

Okay.

And to

But we did do it after the

All right.

That's all I

needed to know.

19

MS. MURPHY:

20

THE COURT:

Thank you.
I will get you my rulings later

21

today, but I do want to bring the jury in and pick up where

22

we left off.

23

return to the stand.

We'll get the jury and ask Dr. Pazzani to

24

(The jury entered the courtroom.)

25

THE COURT:

Welcome back.

Mr. Nelson, you may

Pazzani - direct
1

772

continue.

2

MR. NELSON:

Thank you, your Honor.

3

May I get slide 59, please.

4

BY MR. NELSON:

5

Q.

6

estimating parameters of a learning machine element, and it

7

continues on.

8

A.

So let's turn to the next element of claim 1, the

Okay.

9

MR. NELSON:

Can I get slide 60, please?

10

BY MR. NELSON:

11

Q.

12

the '040 patent.

13

A.

14

machine, wherein the parameters define a user model specific

15

to the user and wherein the parameters are estimated in part

16

from the user-specific data files.

17

Q.

18

So let me have slide 61.

And this is the pull out.

So this is Element C of

Can you read that element to the jury?

Yes, I can.

Estimating parameters of a learning

And let's kind of break that apart a little bit here.

19

And so can you read the Court's definitions of

20

some of the claim terms?

21

A.

22

estimating parameters of a learning machine is estimating

23

parameters of the variables of a learning machine.

24

Q.

25

Yes.

A parameter is a value or a weight and

And let me turn to slide 62, please.
And can you sort of illustrate, or explain what

773

Pazzani - direct
1

you're trying to illustrate here.

2

A.

3

have the user-specific data in Kansas.

4

like

5

and you're trying to create the link profile or the

6
7

Yes.

So in this case, what we're trying to do is, we

from that.

These are things

So here we can see what the profile is.

Maybe we can zoom in a little bit, but this is the

8
9
10

So what we're trying to do is figure out which

11

categories of interest to the user, like living things or,

12

et cetera.

13

Q.

14

A.

Thanks.

And what are the IDs there on this slide?

15
16
17
18
19
20

Q.

And so this PTX-34, this is a Google document; is

21

that correct?

22

A.

23

that large stack that was blocking your view before.

24

Q.

25

we zoom in -- well, first of all, can you explain what this

Yes.

That's part of the

And let's turn to the next slide, please.

,

And can

774

Pazzani - direct
1

is?

2

A.

3

going to tell you about five different learning machines:

4

wink, dilip, rephil, and then two that weren't, those are

Yes.

So this is a different profile.

So, again, I'm

5
6

, your
session profiles.

7

And this is just one other one.

8

the rephil profile.

9

associated with

This is

And now if we zoom in, we can see it's
and that there's

10
11
12
13

Q.

So is there a parameter shown on this slide?

14

A.

Yes.

15

here, or

So this one shows the parameter is
et cetera.

16
17

Q.

Is that parameter estimated?

18

A.

19

Q.

And can you explain, can you explain that?

20

A.

Sure.

21

click on five.

So let's imagine you see ten documents and you

22

MR. VERHOEVEN:

23

THE COURT:

24

MR. VERHOEVEN:

25

Objection.

What's your objection?
The last question and answer and

this question, outside the scope of his report.

775

Pazzani - direct
1

THE COURT:

2

MR. NELSON:

Tell me where it is.
It's in all of the same paragraphs

3

that you referenced before, Your Honor.

4

particulars.

5
6

So it begins at Paragraph 163 and goes on to
paragraph 197.

7

THE COURT:

8

MR. NELSON:

9

Let me find the

Give me a more specific example.
Paragraph 166, beginning estimating

parameters.

10

THE COURT:

11

(Pause.)

12

THE COURT:

13

Right.

I'm reading it.

Thank you.

Mr. Verhoeven, I'm not sure I

understand what you are objecting to.

14

MR. VERHOEVEN:

Thank you, your Honor.

It's

15

very simple.

16

expert answers the question why there's an estimate, which

17

is the question that was posited.

18
19

We don't believe in this paragraph that the

And if you read this, there's no explanation of
that, and we had --

20

THE COURT:

21

MR. VERHOEVEN:

22

THE COURT:

23

As to why it's an estimate?
Yes.

All right.

Do you want to direct me

to somewhere elsewhere he says why this is an estimate?

24

MR. NELSON:

Well, I think it's clear, he's very

25

clear why it's an estimate in here.

776

Pazzani - direct
1
2

THE COURT:

All right.

Is

that right?

3

MR. NELSON:

4

THE COURT:

5

So if it's in 166.

He -I don't want you to get into all the

substance of it, but is that where I look?

6

MR. NELSON:

You can look there, Your Honor.

7

You can also look at his deposition where he was asked this

8

question.

9

THE COURT:

10
11

MR. NELSON:

On page 93, 12 through 25.

We may

need to sidebar on this, too, Your Honor.

12
13

Tell me where.

THE COURT:
you called out first.

All right.
93.

14

MR. NELSON:

15

THE COURT:

16

THE COURT:

What lines, please?

(Pause.)

17

Let me look at the lines

18

12 through 25.
All right.

Thank you.

Mr. Verhoeven, do you stand by your

objection?

19

MR. VERHOEVEN:

To the extent your Honor's rules

20

of the road are if it's in the deposition and not in the

21

report, it's okay, we concede that it's in the deposition,

22

Your Honor.

23

THE COURT:

Okay.

24

MR. VERHOEVEN:

25

road, then we withdraw it.

So if that's the rules of the

Pazzani - direct
THE COURT:

1

Okay.

777

I am going to make those the

2

rules of the road going forward and I do believe that the

3

deposition adequately discloses it and the objection is

4

overruled.

5

BY MR. NELSON:

6

Q.

7

you enlarge

8

jury what one of the parameters is listed there?

9

A.

Yes.

11

Q.

And why is that an estimation or an estimate?

12

A.

Well, what you'd like to have is a very accurate

13

representation of a probability.

14

examples, you might be able to figure out what that

15

probability is, but if you just see three or four examples,

16

it's really hard to get the probabilities right.

Okay.

So let's go back to where we were here.

Can

And so can you again tell the

10

17

If you saw a million

For instance, my mother had four children,

18

three boys, one girl.

Three-quarters of them are boys, so

19

you might assume three-quarters of all people are boys.

20

that's not the case.

21

world, you might get the 50/50.

22

about 51 or 50.5 to 49.5.

23

amount of data and try to extrapolate to the large amount of

24

data, it's just an estimate.

25

Q.

But

If you look at all the people in the
And I actually think it's

But any time you look at a small

Let me ask a question again slightly differently.

778

Pazzani - direct
1

Why is this an estimate of a parameter?

2

A.

Well, the --

7

Q.

So it's -- is it a rough calculation of how

8

interested

9

A.

3
4
5
6

?

Well, in some ways, it's an extremely precise

10

calculation, but that precise calculation is just an

11

approximation of a probability.

12

Q.

13

could we get this one blown up?

And let's go back to the previous slide, 63.

14

And this is the link profile.

15

A.

Well, here,

19

Q.

And can you explain how

20

estimate of that parameter?

21

A.

22

out those things that

And what is the

parameter in the link profile?

16

And

17
18

23
24
25

Yes.

is an

So from -- what we're trying to do is figure

779

Pazzani - direct
1
2
3
4
5
6
7

Q.

And so how are these category interests represented

8

in Google's system?

9

A.

Well, there is a number that is associated with the

16

Q.

And can you explain a little bit more how the

17

categories are chosen to be contained or not contained in

18

the link profile?

19

A.

20

in a minute, that estimates the user's interest in that

21

category, and then all of those things in which the user is

22

20 percent or more interested in get put in that category.

23

So there's some threshold, and once you achieve that

24

threshold, it gets stored in your profile.

25

Q.

10
11
12
13
14
15

Well, there's code, and I think we're going to see it

And let's turn to slide 65, please.

And so can you

Pazzani - direct

780

1

read the Court's definition of learning machine and then of

2

user model specific to the user?

3

A.

4

function or model used to make a prediction, that attempts

5

to improve its predictive ability over time by altering the

6

values or weights given to its variables, depending on a

7

variety of knowledge sources, including monitored user

8

interactions with data and a set of documents associated

9

with the user.

Yes.

So a learning machine is a mathematical

10

Q.

And can you read the definition of "user model

11

specific to the user?"

12

A.

13

implementation of a learning machine updated in part by data

14

specific to the user."

Yes.

15

"A user model specific to the user" is "an

MR. NELSON:

And let me have slide 67, please.

16

BY MR. NELSON:

17

Q.

18

you found for Google Search?

19

A.

20

mathematical models:

And so can you identify the learning machines that

Yes.

21
22
23
24
25

So the learning machines are these abstract

It's the link profiler plus part of the Kaltix
Twiddler.

Or,
It's the dilip and it's the dilip profiler and

the Kaltix twiddler.
And the rephil profiler and the Kaltix twiddler.

Pazzani - direct

There is a Category NavBoost profiler and the

1
2

Kaltix twiddler.
And a session category profiler and the Kaltix

3
4

781

twiddler.

5
6
7
8
9
10

Q.

And just explain how they relate to the graphic there

11

on the lower left.

12

A.

13

the part over here, as the profiler.

14

creates the weights on the variables.

The lower left depicts the abstract learning machine,
That's the thing that

Then the Kaltix Twiddler is the thing that

15
16
17
18
19
20
21
22

Q.

And can you explain to the jury what the user model

23

specific to the user is in Google's system or the user

24

models specific to the user is in Google's system?

25

A.

Yes.

So once there is user specific data, the

782

Pazzani - direct
1

learning machine can operate.

2

called a user model.

3

is, where the user model is the link profile, plus this

4

information, or the dilip profile plus this (indicating).

5
6

And then it creates what is

So in this case, the learning machine

So, in essence, it become substantiated.
no longer an abstract mathematical model.

It's

It's

7
8

Q.

What makes it

?

9

A.

10

Q.

And is it specific to

at that point?

11

A.

Yes, it's

And that is why there

12

are two binders:

13
14
15

Q.

And let's talk now about each of these in each of

16

these five in a little bit of detail.

17

A.

Okay.

18

MR. NELSON:

Let's first put up PTX-222.

19

the next slide.

20

BY MR. NELSON:

21

Q.

22

the day.

23

machine is in Google's system?

24

A.

25

training is the part circled in red.

It's

This is the drawing that we referred to earlier in
Can you identify on the drawing what the learning

Well, the part of the learning machine that does the
So it takes the user's

783

Pazzani - direct
1

history from Kansas, makes some extractions from it, learns

2

some parameters, and then it stores the profile back in

3

Kansas.

4

profiles.

5

Q.

And what about the short time profiles?

6

A.

The short term profiles operate very similarly but

7

there is no need to store them in Kansas.

8

around for a shorter period of time.

9

Q.

That's for three of the profilers.

The long term

They just stay

And let's turn to some Google documents to discuss

10

this further.

11

A.

I can.

12

Q.

Can you identify that document?

13

A.

Yes.

14

infrastructure.

15
16

Can you look in your notebook to PTX-770?

This is the profiler, profile factory
It's a Google document, yes.

MR. NELSON:

I'd like to move PTX-770 into

Federal Defender.

17

MR. VERHOEVEN:

18

THE COURT:

19

(PTX-770 is admitted into evidence.)

20

MR. NELSON:

21

Q.

24

profilers?

25

A.

Can you put up the next slide,

BY MR. NELSON:

23

It's admitted.

please?

22

No objection, Your Honor.

Can you tell me what this document says about

Well, this is a general description of profilers.

784

Pazzani - direct
1

And I think the underlying part is the important part.
It says we use a number of profiles -- profilers

2
3

to compute profiles for a given user.

So there is not

4

just one, there is five different profiles of the user

5

categorizing documents in different ways, or categorizing

6

long and short term interests.

7
8
9
10

Q.

Let me have you turn to PTX-376.

11

A.

(Witness complies.)

12

Q.

Can you identify that document?

13

A.

Yes, I can.

14

Components Used in P Search.

15

Personalized Search.

It is a Google document entitled Generic
P Search stands for

16

MR. NELSON:

I'd like to move 376 into evidence.

17

MR. VERHOEVEN:

18

THE COURT:

19

(PTX-376 is admitted into evidence.)

20

MR. NELSON:

No objection, Your Honor.

It's admitted.

Can you put up slide 70, please?

21

BY MR. NELSON:

22

Q.

23

profilers?

24

A.

25

profilers, and they have several components.

And can you tell me what this document says about

Yes.

This document describes in general the
One component

785

Pazzani - direct
1

looks over the items associated with the user.

2

instance, these are the clicked URL or the documents the

3

user has clicked on.
Then it looks up

4

For

for those.

For

5

instance, it finds the link categories associated with those

6

documents.
Then it aggregates those

7
8

.

It forms

some weighted combination of the

9
10
11
12

Then it writes the profiler back into Kansas in

13

some known form like the link profile we saw on the screen.

14

Q.

And are the profilers -- let me start over here.
MR. NELSON:

15

Let's turn to the next slide.

16

Let's first talk about, are these the five different

17

profiles we're going to discuss?

18

A.

Yes, they are.

19

Q.

Let's first talk about the link profile.

20

look at PTX-30 in your binder?

21

A.

Yes, I can.

22

Q.

And what is that?

23

A.

I don't think PTX-30 is in my binder.

24

been admitted before.

25

not counting well.

Oh, I found it.

Can you

Perhaps it has

I'm sorry.

I'm just

Pazzani - direct

786

1

Q.

Okay.

2

A.

Yes.

3

Q.

Can you tell me what PTX-30 is?

4

A.

PTX-30 is Google's Supplemental Objections and

5

Responses to Plaintiff's Fourth Set of Interrogatories to

6

Google.

7

questions.

Yes.

So it's essentially Google's answers to PUM's

8
9

There is a lot of things in those binders.

MR. NELSON:

I'd like to move PTX-30 into

evidence I.

10

MR. VERHOEVEN:

11

THE COURT:

12

(PTX-30 is admitted into evidence.)

13

MR. NELSON:

14

Q.

16

A.

18

Can you put up slide 72, please.

24 is?

17

It's admitted.

BY MR. NELSON:

15

No objection.

profile is

Can you tell me what Google's answer to our Question

Yes.

It says:

The data used to derive the link

19
20
21
22

I basically have been saying the same thing over
the past hour.
And it says:

23
24
25

And it goes on to say that Google uses a

Pazzani - direct

787

1

Bayesian normalization function, which is really just a way

2

of doing Bayesian estimation.
MR. NELSON:

3
4

slide, please.

5

BY MR. NELSON:

6

Q.

7

profile.

8

A.

9

And let me have you put up the next

the link profile if

Tell me what Glen Jeh said about the link profiler or

Yes.

So Jeh said that the category will appear in

10
11
I described earlier that threshold and whether

12
13

you get above that or not.

14

Q.

15

these issues?

16

A.

17

profiler, for instance.

18

Q.

19

what that is.

20

A.

21

BMRC Katz profiler.

22

inside the computer code but it's not quite the computer

23

code itself.

24
25

And did you look at Google's source code regarding

Yes, I looked at the source code through the link

Let me direct your attention to PTX-76 and ask you

PTX-76 is a header file for the link profiler.
The header file describes what is

MR. NELSON:
PTX-76.

It's

And let me ask the Court to admit

788

Pazzani - direct
1

MR. VERHOEVEN:

2

THE COURT:

3

(PTX-76 is admitted into evidence.)

4

No objection.

It's admitted.

MR. NELSON:

Can you put PTX-76 on the

5

screen, please?

6

BY MR. NELSON:

7

Q.

Can you tell me what this tells you about --

8

A.

Yes.

9

Q.

-- the profile?

10

A.

Yes.

11
12

So these -THE COURT:

Dr. Pazzani, just another reminder.

Wait to answer the question until it's done.

13

THE WITNESS:

14

THE COURT:

15

answer the question when it's done.

16

I want to go home.

No comment on that, but go ahead and

THE WITNESS:

Sure.

17

BY THE WITNESS:

18

A.

19

into this file.

20

used for the programmer to describe what it does.

21

when other programmers look at the file, they can understand

22

it.

23
24
25

So these are the comments that a programmer has put
The comments start with "//" and they're
So that

789

Pazzani - direct
1
2
3
4
MR. NELSON:

5

And let me turn to the next slide.

6

We've already done a little bit of this.

7

BY MR. NELSON:

8

Q.

And so can you explain what this shows?

9

A.

Yes.

10

This is PTX-33.

So these are the top interests in two different

associated with two different Gaia IDs.

One is the

11
12
13
14
15
16

Q.

So are these basically pictures of

17

from Google's computers?

18

A.

19

printed them out for us, and we put them in those binders.

20

Q.

21

Can you turn in your exhibit book to PTX-379?

22

A.

Yes.

23

Q.

And tell me what that is.

24

A.

That is a letter from Quinn Emanuel, Google's

25

attorneys to I guess SNR Denton, PUM's attorneys.

Yes.

So we obtain these from Google, and they

Let's talk about the next profile, the dilip profile.

Pazzani - direct
MR. NELSON:

1
2

I'd like to move PTX-379 be

admitted.

3
4

790

MR. VERHOEVEN:

Your Honor, this also subject to

this morning.

5

THE COURT:

What we discussed.

6

MR. VERHOEVEN:

We have not jet seen the revised

7

version, but subject to it being revised according to the

8

way it was set up this morning, we have no objection.

9
10

THE COURT:

You are working on that revision;

correct?

11

MR. NELSON:

12

THE COURT:

13

Correct, Your Honor.

revision we discussed.

It's admitted subject to that

14

(PTX-379 is admitted into evidence.)

15

MR. NELSON:

And can you put up slide 77,

16

please?

And blow it up.

17

BY MR. NELSON:

18

Q.

19

used to create the link profile?

20

A.

21

actually for the Gaia IDs, and it says it's in the

And can you tell me what this says about the code

This code, this comment describes the dilip profile

22
23

Q.

And did you look at that code?

24

A.

Yes, I did.

25

Q.

Can you turn in your notebook to PTX-98?

791

Pazzani - direct
1

A.

Yes, I can.
Yes, now I have it.

2
3

Q.

And can you tell me what PTX-98 is?

4

A.

PTX-98 is the

5
6
7
8

MR. NELSON:

I move that PTX-98 be admitted into

evidence.

9

MR. VERHOEVEN:

No objection, Your Honor.

10

THE COURT:

It's admitted.

11

(PTX-98 is admitted into evidence.)

12

MR. NELSON:

Can you put the next slide on the

13

board.

14

BY MR. NELSON:

15

Q.

16

dilip profile?

17

A.

18

not going to go into a lot of detail, but, for instance, it

19

contains things like

And is this the code that you just identified for the

Yes, this is just a little bit of source code.

I'm

20
21
22

MR. NELSON:

Let me just put up PTX-25.

23

has already been admitted.

24

BY MR. NELSON:

25

Q.

And the next slide.

Is this an example of a dilip profile?

This

792

Pazzani - direct
1

A.

2

personalization team, Uygar listed his dilip profile.

3

you can see that he is interested in things like references,

4

Wikipedia and about.com, and he is also interested at the

5

bottom in the programming language Java, which is a good

6

thing for a Google employee to be interested in.

7

Yes.

So internal Google documents used by the

MR. NELSON:

8

Q.

And can you turn to slide 80?

BY MR. NELSON:

9

And

And this is Mr. Haveliwala from yesterday.

What did

10

Mr. Haveliwala say about the creation of the dilip profile?

11

A.

12

the user's dilip profile have weights?

So did the different dilip categories that were in

13

Yes.

14

And what did those weights represent?

15
16
17
18

Then the question was:

So I mean, generally the

19
20

And he replied:

21

MR. NELSON:

22

Yes.

And let's turn next to the rephil

profile, slide 81.

23

And can you turn in your notebook to PTX-30?

24

Actually that has already been admitted.

25

just put up slide 82, please.

Let's

Pazzani - direct

793

1

BY MR. NELSON:

2

Q.

3

what creates the rephil profile?

4

A.

Well, the rephil profile has

Q.

And with respect to both the link or all of the link,

Can you tell me what Google said with relation to

5
6
7
8
9
10

dilip and rephil profiles, does the information contained in

11

those profiles change over time?

12

A.

13
14
15
16
17
18

Q.

And what is the purpose of changing this stuff, these

19

weights and values over time?

20

A.

21

by using more data.

22

Q.

23

accurate picture of the user's interest?

24

A.

25

the next time the Google user uses the system, it predicts

To form a more accurate model of the user's interest

And what does the Google system do with that more

Well, it updates and stores it back in Kansas.

And

Pazzani - direct
1

probabilities more accurately.

2

Q.

3

profile?

4

A.

Yes, I did.

5

Q.

Can you look at PTX-69?

6

A.

Okay.

7

Q.

And identify that, please?

8

A.

PTX-69 is Google source code.

9

for

794

Did you look at the code that creates the rephil

.

10
11

It's the source code

MR. NELSON:

I'd like to move Exhibit 69 into

evidence.

12

MR. VERHOEVEN:

No objection, Your Honor.

13

THE COURT:

14

(PTX-69 is admitted into evidence.)

15

MR. NELSON:

It's admitted.

Please put up slide 83.

16

BY MR. NELSON:

17

Q.

18

discusses.

19

A.

20

declaring, if you like, how the rephil profiler works.

21

There is a general way of doing profiles, and this is

22

specific to rephil.

23
24
25

And tell me a little bit about what this code

Well, this code is discussing how the rephil -- or

Pazzani - direct

795

1
2
3
4
MR. NELSON:

5

And let me put up slide 84, please.

6

BY MR. NELSON:

7

Q.

8

does this slide represent?

9

A.

So this is a portion of PTX-34, GGL-PUM0119002.

What

Again, this is the rephil profile that is stored in

10

the

associated with the Gaia ID of this

11

account.

12

Q.

And let me turn your attention to PTX-37.

13

A.

Okay.

14
15

PTX-37 is a Google e-mail.
MR. NELSON:

I move that PTX-37 be admitted into

evidence.

16

MR. VERHOEVEN:

No objection.

17

THE COURT:

18

(PTX-37 is admitted into evidence.)

19

MR. NELSON:

It's admitted.

Can you put up PTX-37, please?

20

BY MR. NELSON:

21

Q.

Can you tell me what this e-mail is?

22

A.

Oh, yes.

23

the Google engineers.

24

search personalization team.

25

Horling's personal profile he mailed to others on the Google

This e-mail is from Bryan Horling, one of
He was head of the search engine, the
And this is actually Bryan

Pazzani - direct

796

1

Search team just to give them an example of what it could

2

do.
It shows, for example, that

3
4
5
6
7
8

Q.

And let me have slide 86, please?

9

A.

And then the numbers there represent

10
11
MR. NELSON:

12

Let me have slide 86, please.

13

BY MR. NELSON:

14

Q.

15

category NavBoost and session.

16

A.

17

What we've been discussing so far,

So let's turn about the two other categories here,

Yes.

Can you explain those?

Category NavBoost is a short-term profile.

18
19

So, for example, when I arrived in

20

Wilmington recently, I was interested in getting a cheese

21

steak.

22

little bit about my interest just that day over the next two

23

hours.

24

few weeks, it gets into my long-term profile, but I have a

25

short-term profile that represents my temporary, my

I typed "cheese steak" into Google and learned a

If I typed cheese steak every day, every week for a

Pazzani - direct

797

1

short-term interests in the last two hours.

2

Q.

3

profile change over that two-hour period?

4

A.

5

a little bit about you, that you like cheese steak.

6

Perhaps with the second query it learns that I like spicy

7

food and it wants to send me to a place that has spicy

8

cheese steak.

9

Q.

And do the values and weights in the short-term

Yes.

So with the first query you type, I learned

And what's the purpose within that

of

10

keeping that information and using it?

11

A.

12

little bit about what your needs are just right then and

13

there, but then it actually does forget about those unless

14

you keep doing it.

15

cheese steak.

16

steak pizza.

17

Q.

18

attempt to improve the predictive ability, to improve the

19

predictive ability by changing the weight?

20

A.

21

search results, it learns a little bit more about you than

22

if you just had two queries and two search results.

23

Q.

24

don't we put that one back up on the board and the next one.

25

The purpose of keeping that information is to learn a

But that's okay.

Today I don't want a

I want a pizza and I don't want a cheese

And during this session window, does Google System

Definitely so.

If you see five queries and five

And let me turn your attention back to PTX-30.

Why

And what does this tell you about the category

798

Pazzani - direct
1

NavBoost profile and how it's created?

2

A.

3

about a particular cookie ID in Kansas,

Okay.

This says, by temporarily storing information

4
5

Google uses this

6

information to infer link categories or dilip category

7

cluster preferences for that cookie in a manner similar to

8

that described above, where above it described the long term

9

profiles link and dilip.

10

Q.

And let me turn to the next slide.

And this is again

11

part of PTX-30.

12

PTX-30, pages 12 and 13, say.

13

A.

14

profiler is in a file called

15

then it goes on to describe things like look up the session

16

queries categories that help implement that.

17

Q.

18

already admitted.

19

please.

What code is used to -- well, tell me what

So it says that the code that implements this

And let me turn your attention to PTX-38.

20

, and

It's

You can just put up the next slide,

And did Google provide an example of a

21

short-term profile?

22

A.

23

after typing one query, Google has learned a little bit

24

about the user.

25

Boston.

Yes.

This was the example we saw earlier, where just

That this user right now is interested in

Perhaps they're visiting.

799

Pazzani - direct
1

Q.

And what -- so let's talk next about the Session

2

Category profile.

3

What does that mean?

4

A.

5

uses the rephil categories instead of the link or dilip

6

categories.

7

Q.

8

more of Google's answers to PUM's questions.

And you see it says rephil below it.

Really, it's the same as category NavBoost except it

And let's turn to the next slide, 91.

9

And this is

What did Google say about the creation of a

10

Session Category profile?

11

A.

It says,

16

Q.

And for each of these linked dilip, rephil, category

17

NavBoost and category session profilers as well as the

18

respective Kaltix twiddler, are they a mathematical model or

19

function?

20

A.

21

estimates the parameters of a learning machine.

22

Q.

And do they do so based on user-specific data?

23

A.

Yes.

24

Gaia ID or

25

Q.

12
13
14
15

Yes.

They are a mathematical model or function that

They get the data from Kansas specific to that

All right.

And in doing so, is the user model the

800

Pazzani - direct
1

implementation of a learning machine, is that made specific

2

to the user?

3

A.

Yes.

7

Q.

And let's turn to slide 92, please.

8

Mr. Horling say about the session, category NavBoost and

9

session profilers -- profiles?

4
5
6
And what did

A.

He said that

15

Q.

Let's turn to the next slide, please.

16

talk about the Search Ads system.

17

A.

Okay.

18

Q.

And let me direct your attention to PTX-112, and in

19

your notebooks?

10
11
12
13
14

And let's now

And can you identify PTX-112?

20
21

A.

Yes.

22

entitled UBAQ in 15 minutes, user-based.

23
24
25

PTX-112 is a document, a Google document,

MR. NELSON:

I want to move PTX-112 into

evidence.
MR. VERHOEVEN:

No objection.

Pazzani - direct
1

THE COURT:

2

801

(PTX-112 was admitted into evidence.)

3

BY MR. NELSON:

4

Q.

5

It's admitted.

A.

What are

?

6
7
8
9
10

MR. NELSON:

And let me have the next slide,

11

slide 96.

12

BY MR. NELSON:

13

Q.

14

in the Search Ads product?

15

A.

16

UBAQ profiler together with the SmartAds System.

17

general system that computes the probability the -- the

18

probability the user will click on an ad, PCTR.

19

Q.

20

user in the Search Ads product?

21

A.

22

UBAQ profile.

23

the user, it knows, for instance,

24
25

And what is the learning machine in the Search Ads,

The learning machine in the Search Ads product is the
That's the

And what is the, what is the model specific to the

That would be the profiler plus SmartAds plus the
So after it has learned a little bit about

Pazzani - direct
Yes.

802

1

Q.

2

the graphic on the left, can you just explain what that

3

graphic is intended to represent?

4

A.

5

the abstract learning machine and then the one on the right

6

that I'd love to see is, is the user model specific to the

7

user.

And there's where you can see that

12

Q.

And how is

13

predictability?

14

A.

Yes.

And we can blow up the bottom part here.

So on

If I -- actually, the graphic on the left is

8
9
10
11
in this, in improving

15
16
17
18
19
20
21
22

Q.

And let me have you look at PTX-397 in your

23

notebooks.

24

A.

25

user-based ads quality.

Yes.

Yes.

397 is a Google document entitled

803

Pazzani - direct
MR. NELSON:

1
2

I'd like to move Exhibit 397 into

evidence.

3

MR. VERHOEVEN:

4

THE COURT:

5

No objection, Your Honor.

It's admitted.

(PTX-397 was admitted into evidence.)

6

MR. NELSON:

Can you put up slide 97, please.

7

BY MR. NELSON:

8

Q.

And what is this document showing?

9

A.

This shows how the

10

is calculated by

Google.

11
12
13
14
15
16
17
18
19

Q.

And what is the parameter in this case?

20

A.

The parameter is the

21

an

22

Q.

And how is that estimated?

23

A.

It's estimated from a small amount of data.

24

you've seen thousands and thousands of ads,

25

.

There's

So if

Pazzani - direct

804

1
2
3

Q.

And does that estimate change over time?

4

A.

Yes, it does.

5

Q.

Okay.

6

A.

Yes.

8

Q.

Now let me turn your attention to PTX-398 in your

9

notebooks.

It moves from one bucket to another?

7

10

A.

Okay.

11

Q.

Can you identify that?

12

A.

Yes, I can.

398 is an e-mail between Google and --

13

MR. NELSON:

I'd like to move 398 into evidence.

14

MR. VERHOEVEN:

15

MR. NELSON:

No objection, Your Honor.

Can you put up the next slide,

16

please?

17

BY MR. NELSON:

18

Q.

What does this document tell you?

19

A.

It basically describes how the

20

done in Google and it says it's based on the

is

21
22

So if you run the profile and it uses the

23

same profile that's already there, it does not update, but

24

if the profile is changed, then it updates.

25

Q.

And the

805

Pazzani - direct
1
2
A.

Yes.

5

Q.

And together is this system a mathematical function

6

or model?

7

A.

8

instantiated for the user.

9

Q.

3
4

Yes.

There's a mathematical model that gets

And let me direct your attention to PTX-869.

Go

10

ahead and look at that in your notebook and tell me what it

11

is.

12

A.

13

Google code that implements the

869 is the

It's

14
15
16

MR. NELSON:

I'd like to move PTX-869 into

evidence.

17

MR. VERHOEVEN:

18

THE COURT:

19

No objection, Your Honor.

It's admitted.

(PTX-869 was admitted into evidence.)

20

MR. NELSON:

Can you put up the next slide,

21

please?

22

BY MR. NELSON:

23

Q.

Can you describe the code?

24

A.

Yes.

25

instance, it says that they'll look at your

So this is portions of the code.

So, for

806

Pazzani - direct
in order to use -- user data in computing the

1
2

profile.

And then it says it looks at the

3
4

Q.

And let me direct your attention to a different

5

portion of PTX.

6

attention to the next slide.

7

That's part of the same exhibit, page 478.

8
9

So that is 100474.

Let me direct your

100, please.

Pull that up.

What does this portion of the code talk about?
A.

10
11
12
13
14

Q.

And let me direct your attention in your notebook --

15

let me have you put up the next slide, PTX-113.

16

user-based ad quality.

17

MR. NELSON:

This is

Can you go back on the animation?

18

Next one.

19

BY MR. NELSON:

20

Q.

Can you just explain what this drawing is?

21

A.

Yes.

22

document.

23

It was within the document itself.

24
25

This is a drawing contained within a Google

It is a hand-done drawing.

Not sure who drew it.

But basically this looks -- this describes
Kansas, which is the database, and then there's various --

807

Pazzani - direct
1

well, the most important part here is this,
So that shows how the profile is stored

2
3

in Kansas, associated with the domain eBay.com, and it's

4
5

Q.

And is there then a user shown on this document?

6

A.

Yes.

7

Q.

And this drawing, this is a Google drawing; right?

8

This isn't something we did?

9

A.

That's right.

10

Q.

Let me -- so the

11

So this is the data for user 1 (indicating.)

This is a Google drawing.
, the

in the Smart Ads System or the Search Ad

12

System, how does this become user specific?

13

A.

Well, it becomes user specific when it looks at the

17

Q.

Let's turn next to the Content Ad System now,

18

slide 102, and why don't we just jump ahead to slide 104.

14
15
16

19

What is the learning machine in the Content Ad

20

system?

21

A.

22

didn't hear a specific name for the profiler, so I will

23

just call it the Content Ads profiler.

24

SmartAds that predicts the, that looks at the user specific

25

portion of -- of the data, of the user data from Kansas

So the learning machine in the Content Ad system, I

And the parts of

808

Pazzani - direct
1

takes that into account in computing the estimated

2

click-through rate.

3

Q.

And so what is the user model specific to the user?

4

A.

The user model specific to the user is instantiated

5

with the user data.

These are a set of categories,

6

These are things that have been

7

in the ads that you've seen recently.

8

Q.

9

it.

10

And why don't we just zoom in on the right side of
Can you explain this further?

A.

So here they're just numbers, if you like, and these

16

Q.

And the face in there, what's that intended to

17

represent?

18

A.

19

user who likes animals or animal ads.

20

Q.

21

predictability over time?

22

A.

Yes, it does.

23

Q.

And how does it do that?

24

A.

It does that by actually learning a profile with

25

each -- with each -- each time it's about to show you an ad,

11
12
13
14
15

The face is a model of a user.

In this case, it's a

And does the Content Ad system attempt to improve its

809

Pazzani - direct
1
2
3
4

Q.

And do those short-term phil clusters have weights

5

associated with them?

6

A.

7
8

Q.

9

A.

10
11

Q.

12

A.

13

Q.

14

And is it, when it's updated with the different
where does that update come from?

Does it come

15

from Kansas again?

16

A.

17

but in the short-term memory of Google, where the short-term

18

profile is.

19

Q.

20

involves a portion of the SmartAds system?

21

A.

22

well.

23

Q.

24

SmartAds system a mathematical function for model?

25

A.

I believe the session data is stored not in Kansas,

And is this also, in the Content Ad system also

Yes.

SmartAds does the auction for Content Ads as

And together, is the profiler and profile and

Yes.

810

Pazzani - direct
1

Q.

And let's take a look at PTX-223, the next slide,

2

please.

3

in the Content Ad system?

4

A.

5

that CUBAQ builds a profile from the person's history, and

6

again that user profile is related

And what does this slide say about the user profile

This slide shows cue back, the profiler -- it shows

7
8
9
10

Q.

Let me ask you to go back to the last slide, 104 for

a minute.

11

And can you zoom in on the right side?
And what are the parameters here?

12

A.

13

rephil clusters that were derived from the ads.

14

Q.

And are they, are they estimated?

15

A.

Definitely.

16

Q.

And how is that?

17

A.

Well, you only look at a small amount of data, so you

18

don't have accurate models of user's interest.

19

sample, a small sample.

20

Q.

21

parameter, what does that represent?

22

A.

23
24
25

Q.

The parameters are the weights associated with the

It's just a

And so the number that is the weight of the

811

Pazzani - direct
1

A.

That's right.

2

MR. NELSON:

3

BY MR. NELSON:

4

Q.

5

yesterday.

6

system?

7

A.

8

Could I get slide 106, please.

they're

And let's talk -- this is Mr. Zamir who spoke
What did Mr. Zamir say about the content ad

So he described the user cookie-based signals, that

9

or he was asked that.

10

if that was the case.

11

short-term rephil clusters.

12

Q.

13

He answered

actually, go to slide 107, please.

14

And then you asked if they were
And he said, yes, they were.

Let me direct your attention in the notebooks --

And what does this portion of PTX-404 show?

15

A.

This is a document that describes the way Google

16

calculates weight on these

17
18
19
20
21

Q.

And let me turn to slide 108, please.

And can you

22

please summarize your opinion whether Google performs each

23

of the Google Search Ads, search and Content Ads YouTube

24

System meets Element 1(c), estimating parameters of a

25

learning machine wherein the parameters are estimated to

812

Pazzani - direct
1

create a user model based on, in part on user-specific data

2

files?

3

MR. VERHOEVEN:

4

now the afternoon.

5

yesterday, this morning.

6

Objection, Your Honor.

It is

to object to the slide.

7
8

We were informed this would be fixed

THE COURT:

It's still not fixed.

I'm going

I will overrule the objection.

It

will be corrected, I trust, by tomorrow.

9

MR. NELSON:
It is fixed.

Apparently -- thank you, Your

10

Honor.

11

can load it; otherwise we can just keep going and do it on a

12

break.

13
14

If we can get online for a minute, we

THE COURT:

We'll do it on a break.

You can

answer the question.

15

MR. NELSON:

Let me restate the question.

16

BY MR. NELSON:

17

Q.

18

Search, Search Ads and Content Ads and YouTube practice

19

Element 1(c) of the '040 patent?

20

A.

Yes, they do.

21

Q.

And they practice each aspect of that element;

22

correct?

23

A.

24

my analysis, went into painstaking detail for each of those

25

analyses.

Can you summarize your opinion, whether the Google

That's correct, although there are summaries there in

813

Pazzani - direct
1

Q.

And that's for both the link, dilip, rephil, Category

2

NavBoost and Session Category Profile Search; is that

3

correct?

4

A.

5

the short-term phil clusters and Content Ads.

6

Q.

7

about analyzing a document.

Yes, it is.

Let's turn to the next slide, please.

8
9

And the UBAQ profile of Search Ads and

MR. NELSON:

Let's talk

Can I have slide 100, please, or

110, and the blow up.

10

BY MR. NELSON:

11

Q.

Can you read that element to the jury?

12

A.

Yes.

13

of the document.

14

Q.

15

about the three link, dilip and rephil portions that we've

16

been discussing before.

17

MR. NELSON:

Analyzing a Document D to identify properties

And let me have slide 111.

And let's talk first

May I have slide 112, please, and

18

the pullout.

19

BY MR. NELSON:

20

Q.

21

about?

22

A.

Yes.

23

Q.

This is PTX-202.

24

A.

This document describes the

25

associated with the link profile.

And can you tell me what this, what this document is

that's
And, again, this shows

814

Pazzani - direct
1

that Google

2
3

Q.

And let me have PTX-25, the next slide.

4

And we put

these up earlier.
Is that an example of that analysis?

5
6

A.

Yes.

That web page is in the Dungeons and Dragons

7

category, for example.

8

Q.

9

about this analysis?

And let me have slide 114.

What did Mr. Horling say

This is not a dilip.

10

Mr. Horling say about the dilip analysis?

11

A.

12

What did

categorization called dilip, and

Well, he agreed there was another form of

13
14

Q.

Let me have the next slide, PTX-115.

15

example of a dilip categorization?

16

A.

Yes, it is.

17

Q.

And let me have the next slide, PTX-30.

18

category NavBoost, a portion of Google's answers to our

19

questions.

20

And this is an

And this is

Can you tell me what that says about the

21

creation of the category NavBoost or the analysis in

22

category NavBoost?

23

A.

24

category or dilip cluster preferences.

25

document analysis.

Yes.

It says that category NavBoost uses the link
So it uses that

Pazzani - direct
1

Q.

Let's turn to the next slide.

2

one down.

3

What is PTX-24?

4

A.

5

815

set of interrogatories to Google and their responses.

You can turn in your notebook first to PTX-24.

PTX-24 is another set of interrogatories, the fourth

6
7

Actually, take that

MR. NELSON:

I move that PTX-24 be admitted if

it hasn't already.

8

MR. VERHOEVEN:

9

THE COURT:

No objection.

All right.

10

(PTX-24 was admitted into evidence.)

11

MR. NELSON:

It's admitted.

Can you put up slide 117, please.

12

BY MR. NELSON:

13

Q.

14

documents in the context of rephil?

15

A.

It says that

21

Q.

All right.

22

118.

23

rephil profile?

24

A.

25

associated with this web page.

And what does this say about whether Google analyzes

16
17
18
19
20
And let me turn your -- turn to slide

And is this just an example, this is PTX-25 of the

Yes.

This is an example of a type of rephil profile

816

Pazzani - direct
1

Q.

And let me turn to PTX-30.

2

Session Category analysis.

3

in evidence.

4

And let's talk about the

Just look at slide 119.

This is

What does this tell you about the Session

5

Category analysis?

6

A.

7

general approach is very similar to the rephil long-term

8

category except instead of looking at the long-term data,

9

it's just looking at the short-term data, but certainly

10

Well, I think the most important part is that the

this uses rephil, because it's called

11
12

Q.

And let me have you turn in your notebook to PTX-17.

13

And can you identify that document?

14

A.

15

search overview, or the life of the query.

Yes.

16
17

This document is a Google document called web

MR. NELSON:

I move that PTX-17 be admitted into

evidence.

18

MR. VERHOEVEN:

19

THE COURT:

No objection, Your Honor.

It's admitted.

20

(PTX-17 was admitted into evidence.)

21

MR. NELSON:

Thank you.

22

BY MR. NELSON:

23

Q.

24

well as we just discussed?

25

A.

Pull that up.

Now, does Google analyze documents in other ways as

Yes.

So Google, one of the most important things

817

Pazzani - direct
1

Google does is it indexes the content of what's in

2

individual web pages.

3

finds the words in them, so later on when you type those

4

words as a query, it can find the documents.

5

So it looks at the web pages and

So it stores this information about those pages

6

in an index that maps words or phrases to these documents so

7

Google can find them.

8

Q.

9

of Search Ad.

Now let's talk about the analyzing a document aspect

10

MR. NELSON:

11

have slide 122, please.

12

And let me have slide -- let me

Go ahead and put the other one up.

Blow it up.

13

BY MR. NELSON:

14

Q.

What is up on the top?

15

A.

Up on the top is an example of an ad.

16

Q.

Can you explain the different parts?

17

A.

Well, the most important part here that Google does

18

for analysis in Search Ads is it figures out what the

19

So an ad has a number of

20

components, and one form of analysis is to break things into

21

their constituent components.

22
23
24

Q.

And let me have you turn in your notebook to PTX-402.

25

A.

(Witness complies.)

818

Pazzani - direct
1

Q.

What is PTX-402?

2

A.

PTX-402 is a Google document, Ads Quality.

3

Model Overview.

4

MR. NELSON:

5

SmartAds

And move that Exhibit 402 be

admitted into evidence.

6

MR. VERHOEVEN:

No objection.

7

THE COURT:

8

(PTX-402 is admitted into evidence.)

9

MR. NELSON:

It's admitted.

Can you put up slide 123, please.

10

BY MR. NELSON:

11

Q.

12

documents for Search Ads?

13

A.

14

the SmartAds system.

15

results of this analysis by finding

16

MR. NELSON:

What does this tell you about the analyzing of

Well, there are

that are used in

And this shows that it uses the
.

Let's turn next to the Content Ad

17

system and puts up slide 125, please.

18

BY MR. NELSON:

19

Q.

20

you about whether Content Ads analyzes documents?

21

A.

22

with

23

of rephil.

24

Q.

How does that speak to the analysis of documents?

25

A.

Well, that shows that the ads have been analyzed so

This is Exhibit 403.

Yes.

So what does this document tell

Google admits that Content Ads are associated
phil is the earlier version

Pazzani - direct

819

1

that they can be associated with phil clusters.

2

Q.

Let me have you turn in your notebook to PTX-411.

3

A.

(Witness complies.)

4

Q.

What is PTX-411?

5

A.

PTX-411 is a Google document entitled

6

AdGroupRephilGenerator.

7

Q.

And what is that document?

8
9

MR. NELSON:
Oops.

Can you put up slide 126, please?

Let me move 411 into evidence.

10

MR. VERHOEVEN:

No objection.

11

THE COURT:

12

(PTX-411 is admitted into evidence.)

13

MR. NELSON:

It's admitted.

Put up the next slide, please.

14

BY MR. NELSON:

15

Q.

What does this document tell you about analysis?

16

A.

Well, it says that Google has a tool called the

17
18
19
20

MR. NELSON:

Let me turn to the next slide,

21

please.

22

BY MR. NELSON:

23

Q.

24

Search Ads, Search and Content Ads, and YouTube practice

25

Element 1(d), analyzing a document step?

Can you summarize your opinion whether Google's

820

Pazzani - direct
1

A.

2

Yes, each of them do.
MR. NELSON:

Let's turn to Element 1(e),

3

estimating a probability.

4

slides, please?

5
6

THE WITNESS:

Can you put the claim up on the

I'm not sure if it would be

possible to take a break at this point.

7

THE COURT:

8

THE WITNESS:

9

THE COURT:

10

Sure.

It is possible.

I have to use the restroom.

Not a problem.

Ladies and gentlemen of the jury, we'll take a

11

break.

No talking about the case during the break.

12

get you back shortly.

13

(Jury left courtroom.)

14

THE COURT:

15

(Brief recess taken.)

16

*

17

(Proceedings reconvened after recess.)

18

THE COURT:

19

THE WITNESS:

20

THE COURT:

21

(Jury returned.)

22

THE COURT:

23

are ready to continue.

We'll

24
25

*

Okay.

We will be in recess.

*

Are you okay to continue?
Fine.

We'll bring the jury in.

Ladies and gentlemen of the jury, we

Mr. Nelson, you may proceed.
BY MR. NELSON:

Pazzani - direct

821

1

Q.

Are you feeling better, Dr. Pazzani?

2

A.

Yes, I am.

3

Q.

So let's talk about element 1(e) or step 1(e) of the

4

'040 patent, the estimating of probability step.

5

A.

Well rested.

Sure.

6

MR. NELSON:

Can I get the next slide, please.

7

BY MR. NELSON:

8

Q.

Can you read the jury the claim element 1(e)?

9

A.

It's estimating a probability that an unseen document

10

d is of interest to the user wherein the probability is

11

estimated by applying the identified properties of the

12

document to the learning machine having the parameters

13

defined by the user model.

14

Q.

15

sort of visually demonstrate what this element is about?

16

A.

And did you prepare an animation or illustration to

Yes, I did.

17

MR. NELSON:

Can I get the next slide, please.

18

BY MR. NELSON:

19

Q.

Can you explain?

20

A.

Yes.

21

was created by the learning machine.

22

model has a number of categories, animals, sports, music,

23

computers, and associated with those categories are levels

24

of interest.

25

has a high weight for animals.

So what you see here is a user model, and this
And here, the user

So this particular one is the animal lover and

822

Pazzani - direct

Then there are a set of documents that the user

1
2

may potentially be of interest to him.

And you look at the

3

properties of those documents.

4

them and it estimates the probability that the user would

5

be interested in that in part by looking at the properties

6

of the documents.

7

Q.

8

machine?

9

A.

The user model looks at

In this case, like the topics.

Is that the right-hand portion of the learning

Yes.

The estimating part is in the right-hand

10

portion of the learning machine.

11

before about predicting whether something was an apple.

12

Here, we're predicting whether

13

Q.

14

works at a high level with respect to web pages that the

15

system would have, going back, as part of the search result?

16

A.

17

query.

18

finds documents that contain the words in that query.

19

That's really what we saw

likes it.

In the Google Search system, can you explain how this

Sure.

So at a very high level, the user types a

Google looks in

its document database and

Then it has a collection of possible documents.

20
21
22
23

Then it estimates the probability of interest by

24

looking at each document in the top few and sees if the user

25

would be interested by examining the properties.

So the

823

Pazzani - direct
1

link profiler looks at the link categories of the documents

2

and link profiler of the user.
The rephil profiler does similarly.

3
4

Q.

And then what portion of the code then does the

5

actual comparison of the document profile -- the document

6

properties in the web documents with the profile?

7

A.

8

things:

9

interested in the document, and

That is part of the Kaltix Twiddler.

So it does two

One is it estimates the probability the user is

10
MR. NELSON:

11

Okay.

Now let's break this down

12

into kind of its components parts.

13

please?

14

BY MR. NELSON:

15

Q.

16

would be?

17

A.

18

degree of belief or likelihood."

19

Q.

20

your opinion, did you rely on the Court's construction?

21

A.

22

issued my opinions and I relied on them heavily.

23

Can I get slide 131,

Can you tell me what the Court-defined "probability"

Yes.

The Court defined "probability" as "a numerical

I don't know if I asked you this before, but in forming

Yes.

So the Court construed the claims before I

MR. NELSON:

24

BY MR. NELSON:

25

Q.

So let's turn to slide 132.

Can you explain what this slide is intended to show?

824

Pazzani - direct
1

A.

Yes.

2

thought I would use a blackjack example to describe odds.

3

Odds are another form of likelihood.

4

talking about probabilities between 0 and 1 but odds are

5

just another way of expressing a probability.

6

So since we're here near Atlantic City, I

So far we've been

So imagine you are playing blackjack and you

7

have cards that sum to 12.

You don't want to go over 21 so

8

you don't want to get a 10.

So what are the odds you don't

9

get a 10?

So if you look at the complete deck of cards,

10

there are 36 cards less than 10, and then the face cards and

11

the 10 are equal to 10 in blackjack.

12

16 that you won't get a 10 or 9 to 4 or better than 2 to 1

13

odds in your favor.

14

that card because the orders are 2 to 1 in your favor.

15

Q.

And let me turn to the next slide.

16

A.

The next slide.

17

Q.

What does this slide intend to show?

18

A.

Well, what this slide is intended to show is in

19

blackjack, they don't want you to know the odds precisely,

20

so what they do is they actually use five decks of cards and

21

they have some of the cards they're not going to use and

22

that is so you actually can't calculate the odds.

23

have to estimate them.

24
25

So ou might want to think about drawing

MR. NELSON:
BY MR. NELSON:

So your odds are 36 to

You just

Now, let's turn to the next slide.

825

Pazzani - direct
1

Q.

And so can you tell me the Court's definition of

2

"estimating?"

3

A.

4

"approximating or roughly calculating."

5

Q.

6

BY MR. NELSON:

7

Q.

And so you can you explain this slide?

8

A.

Yes.

9

want to ask you the question what is the probability you

10

will pull out a yellow M&M when you pull out that box of

11

M&Ms?

12

frequency is 3/16ths, but can you use that to predict what

13

will happen in the next bag of M&Ms?

14

you might have five, you might have two yellow M&Ms.

15

even though you can accurately compute that this is 3/16ths,

16

it's just an estimate what is going to happen in the future.

Yes.

The Court's definition of "estimating" is

And so the next slide, please.

This is a package of M&Ms.

And let's imagine I

So in this case, there are 16 M&Ms and so the

17

MR. NELSON:

You might have four,
So

Can I have the next slide?

18

BY THE WITNESS:

19

A.

20

better but you will never get it perfectly.

21

Q.

22

"estimating a probability (p)u(d) that an unseen document d

23

is of interest to the user u?"

24

A.

25

numerical degree of belief or likelihood that the unseen

If you have a four pound bag of M&Ms, you can do it

So can you tell me the Court's definition of

Yes.

It's "approximating or roughly calculating a

826

Pazzani - direct
1

document d is of interest to the user given the information

2

that is known about the unseen document."

3

Q.

Let's go through a couple of other Court definitions

4

here.

This is the "learning machine" definition we talked

5

about before and the "user model specific to the user"

6

definition.

7

A.

That's correct.

8
9

Correct?

MR. NELSON:
please.

And so let me have slide 138,

Now we're in the applying phase here.

10

BY MR. NELSON:

11

Q.

12

having the parameters defined by the user model in Google's

13

Search?

14

A.

15

we've been talking about, five different ways of estimating

16

the probability that a user is interested in the document.

17

So there is the link profiler, plus the Kaltix twiddler,

18

plus that individual user's link profile.

19

profiler and the Kaltix twiddler or that individual user's

20

dilip profile.

21

session category.

22

specific to that user informed by looking at that user's

23

data.

24

Q.

25

parameters defined by the user model for the link profile is

Can you identify what the different learning machines

Yes.

So in Google Search, there are five profiles

Or, the dilip

Similarly for rephil, category NavBoost and
So each has a different set of parameters

So just for example, a learning machine having the

827

Pazzani - direct
1

the link profiler plus the Kaltix twiddler plus the user's

2

link profile?

3

A.

Yes, I think I said that a couple minutes ago.

4

Q.

Is that what makes it specific to the user?

5

A.

The fact that it has the user's link profile, yes.
MR. NELSON:

6

And let's talk about the individual

7

different ones as quickly as we can here.

8

PTX -- or slide 139, please.

9

Let me have

BY MR. NELSON:

10

Q.

And what does this tell you about what is doing the

11

estimating here?

12

A.

13

NavBoost and Session Category, it's the Kaltix Twiddler that

14

does this estimating.

15

Q.

Well, for the link, dilip and rephil and Category

16
17

A.

18
19
20

Q.

We'll get into that in just a minute.

21

A.

Yes.

22

MR. NELSON:

23

Put that up.

24

BY MR. NELSON:

25

Q.

So let me have slide 140, please.

Can you identify what you are representing here in

828

Pazzani - direct
1

slide 140?

2

A.

3

Personalization system that are highlighted in yellow that

4

perform the function of estimating a probability that the

5

user is interested in the document.

6

Q.

And that is PTX-22?

7

A.

Yes.

Yes.

MR. NELSON:

8
9

So these are the portions of the Google Search

Let me have the next slide, please.

Well, let me go back to the first one.

10

BY MR. NELSON:

11

Q.

12

figure from page 465, 345 of PTX-22 shows the learning

13

machine having, using the parameters used by the user model?

14

A.

15

out of Kansas and goes into the twiddler and the profiler as

16

well, a portion of that.

17

highlighting.

So which portions are there -- which portions of the

It's the portions highlighting the profile that comes

18

MR. NELSON:

I guess that was left out of the

Okay.

Let's turn to slide 141.

19

And let's talk first about these three:

20

Category NavBoost.

21

BY MR. NELSON:

22

Q.

23

next to it?

24

A.

25

but

What does it mean?

link, dilip and

What is

I think we might have mentioned that briefly before,

829

Pazzani - direct
1
MR. NELSON:

2

Let me have PTX-30, slide 143.

3

BY MR. NELSON:

4

Q.

And what is doing that for the link profile?

5

A.

It's part of the

6

the top end documents that have been retrieved and then

7

estimates the probability of the user interest in each of

8

those.

9

Q.

that looks at

Turn to the next page.

10

sorry.

11

A.
Q.
A.

I'm

Tell me what this says?

13

Oops.

Yes.

12

That is 144.

143.

143 is the

14
15
16
17

MR. NELSON:

Let me have 144, please.

18

BY MR. NELSON:

19

Q.

What did Mr. Horling say about this?

20

A.

What are the ranking algorithms used by the Kaltix

21

Twiddler?

22
23
24
25

There is one that corresponds to the dilip and
link profiles.
What is the name of that one?
is used for those.

Pazzani - direct
MR. NELSON:

1

830

That is at 159, 8 through 13.

2

BY MR. NELSON:

3

Q.

4

please.

5

A.

433?

6

Q.

Yes.

7

A.

Yes.

8

Q.

And can you tell me what that document is?

9

A.

It's a Google document called personalized search.

Let me have you turn in your notebook to Exhibit 433,

10

MR. NELSON:

11

MR. VERHOEVEN:

12

THE COURT:

13

And can we move 433 into evidence?
No objection, Your Honor.

It's admitted.

(PTX-433 was admitted into evidence.)

14

MR. NELSON:

Slide 145, please.

15

BY MR. NELSON:

16

Q.

17

to do the estimates of probability?

18

A.

19

algorithm called

And so what does this tell you about what's going on

Well, what Brian Horling just mentioned, there's an

20
21

Q.

22

PTX-729, slide 146.

23

A.

24
25

And can you turn your attention -- let me have
And what does this say about

?

Pazzani - direct

831

1
2
3
4
5

Q.

And is this done using logistic regression?

6

A.

Yes.

7

that calculates the log of the odds.

8

Q.

9

mathematical calculations on a white board basically to

It uses what I would call a log linear model

And are you prepared to do sort of -- some

10

further teach what logistic regression is?

11

A.

Yes, I would be happy to.

12

Q.

You may step down.

13
14

(The witness left the witness stand and
approached the easel.)

15

MR. VERHOEVEN:

16

THE COURT:

17

THE WITNESS:

18

Yes, of course.
Is it possible to put the

PowerPoint up also?

19

MR. NELSON:

20

THE WITNESS:

21

MR. VERHOEVEN:

22

THE COURT:

23

Permission to approach?

Sure.
Anywhere is fine.
Sorry.

I can't see.

Let's let them get set up and then

you can relocate.

24

MR. VERHOEVEN:

25

THE WITNESS:

This is fairly heavy.

And high.

832

Pazzani - direct
1

MR. VERHOEVEN:

2

THE WITNESS:

3

MR. NELSON:

I think that

I will just hold this here for a

THE COURT:

That's where you want it,

Mr. Nelson?

8

THE WITNESS:

9

MR. NELSON:

That's fine.
Is that fine with you, Mr. Pazzani?

10

THE WITNESS:

11

THE COURT:

12

All right.

second.

6
7

Okay.

will be fine.

4
5

And high.

Yes.

Mr. Verhoeven, will you be able to

see it from there?

13

MR. VERHOEVEN:

Yes, I will, Your Honor.

14

THE WITNESS:

15

So first I'm going to tell you about log linear

Thanks.

16

models.

17

like this in a few hours in a course.

18

this for two minutes, but first let's talk about logarithms.

19

Do you remember logarithms from high school, maybe?

20
21
22

So this is going to take -- I normally do something
I'm only going to do

So let me give you a few numbers and I will tell
you about logarithms.
First we know about the number ten, the number a

23

hundred, the number a thousand, and I'm also going to do

24

one-tenth and let's do one-thousandth also.

25

Okay?

So those are numbers that you encounter every

833

Pazzani - direct
1

day.

You can rewrite this if you like as ten times ten, or

2

ten squared is the same as a hundred; right?

3

is ten times ten times ten or ten to the third.

4

And a thousand

What you might not know or remember is that

5

one-tenth is actually ten to the minus one and one

6

one-thousandth is ten to the minus three.

7

number 1,000 same as that, but because it's one-thousandth,

8

we do minus three instead.

9

So it's the

Okay.

So that's just a quick refresher on exponents.

10

So now what we can do is figure out what the log of each of

11

these numbers is.

12

Okay.

13

one-tenth is minus one.

14

is minus three.

The log based ten is just the exponent.

So the log ten of a hundred is two.

15

The log ten of

The log ten of one over a thousand

Okay?

Now I want to describe how these might be used

16

to deal with the log of the odds.

17

that over.

18

here.

Normally, I would just flip this over like this

19

MR. NELSON:

20

THE WITNESS:

21

MR. NELSON:

22

THE WITNESS:

23

MR. NELSON:

24

THE WITNESS:

25

If you can just flip

prop this up.

We're having a slight problem here.
All right.

Okay.

Will you help me?
No, but I will watch.
It's falling over.
I will grab this and you try to

834

Pazzani - direct
1

MR. NELSON:

2

for now.

3

this page over.

We lost a leg here.

I will get the other one here.

THE WITNESS:

5

MR. NELSON:

Yes.

That will last for awhile.

It's very heavy.

This isn't going

to work either.

7

All right.

8

THE WITNESS:

9

Why don't you flip

I will get the other one.

4

6

Let's do this

Continue, Dr. Pazzani.
Sure, except I lost my pen.

Let's

get the green out.

10

MR. VERHOEVEN:

While they're setting up, Your

11

Honor, I would request that we proceed not in narrative

12

teaching fashion, but in Q&A.

13

THE COURT:

That's fine.

14

an objection, that's sustained.

15

To the extent that's

Mr. Nelson.

16

MR. NELSON:

Let's ask some questions,

Yes.

17

BY MR. NELSON:

18

Q.

19

how that relates to probabilities?

20

A.

21

now what I'd like to do is describe what the log of the odds

22

are and now that's used to estimate probabilities.

23

Q.

Go ahead.

24

A.

Do you think it's going to stay?

25

going to start with something, what's the probability

Can you explain that the math you described earlier,

Yes.

What I've described so far what logs are and

And I'm actually

835

Pazzani - direct
1

someone gets lung cancer?

2

something like that, but we use these things in medicine

3

very often.

4

Okay.

I don't want to talk about

And let's just assume the probability that

5

one, or the odds that one gets lung cancer is one in a

6

thousand.

7

remember, is the log of that is minus three.

8

to say that's the prior probability or the base.

So one way to represent that, one in a thousand,

9

So I'm going

If we know nothing about you, there's a

10

one-in-a-thousand chance you have lung cancer.

11

Unfortunately, it's actually a little bit higher.

12

doing this to make the numbers kind of round.

13

I'm just

Now, if you smoke, unfortunately, your odds of

14

getting lung cancer increase by a factor of ten.

15

to remember, ten to the one is one way of representing ten,

16

so I would add a one there and I'd have a variable S.

17

variable will be zero if you smoke -- I'm sorry.

18

you don't smoke and one if you smoke.

19

You have

That

Zero if

So now if you are a smoker, this equation

20

produces two, which is the log of the odds or, really, ten

21

to the minus two.

22

So now instead of there being a one in a

23

thousand chance that you smoke, there's -- I'm sorry, that

24

you have cancer, there's a one in a hundred chance that you

25

have cancer because this increased it by a factor of ten.

836

Pazzani - direct
1

Now let me tell you something lucky.

If you

2

live in Hawaii, there's a 50-percent less chance you have

3

lung cancer.

4

that many cars.

5

of half is minus .301.

6

remembered.

7

The air is really clean there.

It's in the middle of the ocean.

The log

That's the only one that I

Okay.
So then we'll have another variable, minus .301

8

times H.

9

your odds decrease.

10

There's not

So if you live in Hawaii, that takes on a one and
Okay.

If you don't live in Hawaii,

that's a zero and then your odds are the same.

11

So if you can combine each of these, and there

12

are many, many more variables you can think about.

13

you've been exposed to asbestos, you're more likely to have

14

cancer.

15

likely to have cancer.

16

likely to have lung cancer.

17

on and on.

18

Q.

19

Google system?

20

A.

21

want to explain one other thing first.

22

If

If you have been exposed to radon, you're more
If your parents smoked, you're more
So this equation can keep going

Can you relate that to the estimated probability in a

Yes.

I will do so in just one minute, but I actually

So these are what we call indicator variables.

23

They're either zero or one.

Either you smoke or you don't.

24

But it turns out that the amount that you smoke affects the

25

probability that you have lung cancer.

If you smoke just a

837

Pazzani - direct
1

little bit, you're less likely to have lung cancer, and if

2

you smoke a lot, you're more likely to have lung cancer.

3

So one way is to have a variable not S that's

4

either zero or one, but to have a variable C that represents

5

on average the number of cigarettes you smoke a day.

6

So now instead of that, the equation might be

7

minus three plus one-twentieth times C.

8

cigarettes in a pack, so if you smoke one pack a day,

9

your odds of getting lung cancer increase by a factor of

10

There are 20

ten.

11

MR. VERHOEVEN:

Excuse me, Your Honor.

12

going to object as a narrative.

13

I'm

the question.

14

THE COURT:

It's not even responsive to

Sustained.

Dr. Pazzani, let's

15

answer the question that counsel asked.

16

BY MR. NELSON:

17

Q.

18

Google Systems?

19

A.

20

predict the probability that you are going to click on a

21

document as to the probability that you might have that one

22

might get lung cancer.

23

like this.

24
25

Can you relate this to the estimated probability of

Exactly.

Let me do that now.

So now we're trying to

So there would be a similar equation

Let's just imagine there's a one-in-ten
chance that one clicks on a document in Google.

So that

838

Pazzani - direct
1

would be, we'd start with minus one, which is ten to the

2

minus one is one-tenth.

3

increase or decrease that.

4

And then there will be factors that

So, for instance, we saw the log linear

5

equation before.

6

categories and your profile and the dilip categories and the

7

document, that increases the chances one will click on it.

8

So let's add a little bit, and I will just say

9
10

it's .2.

If you have overlaps between your dilip

I don't know what this number is.

I will add a

little bit times the number of dilip categories you have.

11

And so that's saying the odds increase of

12

clicking on a document if there's an overlap between the

13

dilip category and the dilip categories in the documents.

14

And there's also an odds that will increase, I will say

15

that's .3.

16

Google in particular, but they don't matter.

17

number is D.

18

Again, I don't know what these numbers are in
The important

That comes from your profile.
THE COURT:

All right.

19

Mr. Nelson.

20

Let me stop you there,

becoming too much of a narrative.

21

I need you to ask questions because this is

MR. NELSON:

22

probably done with this.

23

Okay.

That's fine.

I think we're

Are you done with this

demonstration?

24

THE WITNESS:

I think I've probably illustrated

25

enough, but this is the general idea.

I might have to

839

Pazzani - direct
1

come back and add one or two things perhaps on

2

cross-examination.

3

MR. NELSON:

4

THE COURT:

5

MR. NELSON:

6

Go ahead.

All right.
I'd like to mark this as a

plaintiff's exhibit.

7
8

Thank you.

THE COURT:

As a demonstrative exhibit or

substantive?

9

MR. NELSON:

Well, a demonstrative exhibit, I

10

guess.

Well, a substantive exhibit as well.

11

talked about it and summarized his testimony.

12

THE COURT:

13

MR. VERHOEVEN:

I mean, he has

Mr. Verhoeven?
Your Honor, we would consider

14

this a demonstrative exhibit.

15

being marked as a demonstrative exhibit, but obviously we

16

object to its being admitted.

17

THE COURT:

18

exhibit.

19

We have no objection to it

We'll mark it as a demonstrative

To the extent it's being offered into admission,

it's not being admitted.

20

MR. VERHOEVEN:

21

THE COURT:

22

25

For the record, you'll need to make

up a number for it.

23
24

Thank you, Your Honor.

MR. NELSON:

Oh, let's call it Demonstrative

Exhibit 1.
THE COURT:

Demonstrative Exhibit 1.

Okay.

840

Pazzani - direct

(Demonstrative Exhibit No. 1 was marked for

1
2

identification.)

3

MR. VERHOEVEN:

Your Honor, I apologize.

I just

4

realized none of these slides have been marked with

5

demonstrative numbers either, so we'll need to take care of

6

that.

7
8

THE COURT:

Right.

Okay.

I will leave it to

you all to do that sometime before we're done.

9

MR. NELSON:

We'll do that.

10

BY MR. NELSON:

11

Q.

Let's continue.

12

A.

Can I go back to the prior slide?

13

Q.

Yes.

14

A.

Okay.

15
16

Go back to slide 146, please, PTX-729.
So now I hope with that -THE COURT:

Hold on.

There's no question

pending.

17

Do you have a question, Mr. Nelson?

18

MR. NELSON:

Yes, I do.

19

BY MR. NELSON:

20

Q.

21

to the demonstration that you just did?

22

A.

23

linear model, that's one of those equations that I was

24

drawing out, to predict the long-term probability based on a

25

number of

So can you explain how the

Yes.

relates

So now what you see here is that there's a log

such as

841

Pazzani - direct
1

So each of those will have a multiplier

2
3

effect.

You're more likely to click on it

4
5

You're also more likely to click on it based on other

6

factors, such as the

7

if the document really
.

That's independent of the user.

8

But the two that are underlined are the dependent parts, and

9

those are important parts of estimating the probability that

10

the user is of interest -- that the document is of interest

11

to the user.

12

Q.

13

some of these systems work on aggregate data.

14

explain how this equation relates to those statements?

15

A.

Earlier in I think opening, Google mentioned sort -Can you

Yes.

16
17
18

There are some components of

19

this that are true for all users, how likely you are to

20

click on something if it's in the first position versus the

21

tenth position.

22
23
24
25

Q.

Okay.

And is this used for link dilip and category

842

Pazzani - direct
1

NavBoost?

2

A.

Yes, that's correct.

3

Q.

And let me have you look in your notebook at PTX-200.

4

And can you identify PTX-200?

5

A.

Yes.

PTX-200 is the code, the computer code for the

6
7
8
9
10

MR. NELSON:

I'd like to move PTX-200 into

evidence.

11

MR. VERHOEVEN:

12

THE COURT:

13

No objection, Your Honor.

It's admitted.

(PTX-200 was admitted into evidence.)

14

MR. NELSON:

Let's put up PTX-200 on the board

15

and Mr. Horling's testimony about it.

16

BY MR. NELSON:

17

Q.

18

PTX-200?

19

A.

20

Fortunately, you won't have to see the beginning.

21

at the code on the bottom of line 508.

22

And can you just read what Mr. Horling said about

Yes.

You asked him to turn to 99938 of the document.
And look

And it says,

23
24

MR. NELSON:

25

THE WITNESS:

Can you pull that out of the slide?
I think it's the next -- next

843

Pazzani - direct
1

animation.

2

MR. NELSON:

The next slide.

3

THE WITNESS:

4

So what was being referred to in the questioning

Yes, I think it is the next slide.

5

is these particular lines of code.

Now, I don't want to go

6

through all the details, but you can read the English.

7
8
9

.

10

Q.

And can you look at PTX-97 in your notebook and tell

11

me what that is?

12

A.

PTX-97 is the header file for Kaltix pending servlet.

13
14

MR. NELSON:

I'd like to move PTX-97 into

evidence.

15

MR. VERHOEVEN:

Just one second.

16

(Pause while counsel conferred.)

17

MR. NELSON:

18

MR. VERHOEVEN:

19

MR. NELSON:

20

(Pause.)

21

MR. VERHOEVEN:

22

THE COURT:

The next slide, please.
I'm sorry.

What was the number?

97.

No objection, Your Honor.

It's admitted.

23

(PTX-97 was admitted into evidence.)

24

BY MR. NELSON:

25

Q.

Can you tell me what PTX-97 is?

844

Pazzani - direct
1

A.

Yes.

PTX-97, the part that we're looking at is just

2
3
4

.

We saw a Google document describe this.

is the computer code that describes it.

This

It says,

5
6
7

Q.

8

PTX-382.

9

Okay.

Can I have you look in your notebook at

MR. NELSON:

Can you take the slide down.

10

THE WITNESS:

11

MR. NELSON:

12

MR. VERHOEVEN:

13

THE COURT:

14

Yes.

PTX-382 is a Google e-mail.

I move PTX-382 into evidence.
No objection, Your Honor.

It's admitted.

(PTX-382 was admitted into evidence.)

15

MR. NELSON:

And let's put up slide 151.

The

16

next one, next one, next one.

17

BY MR. NELSON:

18

Q.

What is this one?

19

A.

Well, I think what we're getting at here is what does

20

it mean

And what does this indicate?

21
22
23
24

Q.

And what does Mr. Horling say about this?

25

slide, please.

The next

845

Pazzani - direct
1

A.

2
3
4
5
6
7
8
So what that's really getting at, if there's

9
10
11
12

Q.

And so what is the estimated probability for the

13

respective link dilip and category NavBoost profile that is

14

being estimated in this element?

15

A.

It's the part of the

16
17
18

MR. NELSON:

Let's go to slide 153.

And let's talk about the rephil and session category.

19

And may I have slide 154, please.

It's PTX-154.

20

BY MR. NELSON:

21

Q.

What does this say about the rephil profile?

22

A.

It talks about the rephil profile, and it uses some

23

, and it

24

describes in a little bit more detail about how it works.

25

Fortunately I'm not going to go through the math of this.

Pazzani - direct
1

846

But essentially

2
3
4
5
MR. NELSON:

6

Can I have slide 155, another piece

7

of PTX-24.

8

BY MR. NELSON:

9

Q.

What is this document, sir?

10

A.

This goes into a little bit more detail about rephil,

11

and it says:

12
13
Basically,

14

-- I'm sorry,

15
16
17
18
19
20

So actually if you go to the Alvarado example

21

that Konig used earlier, his wife was interested in

22

elementary schools.

23
24
25

Q.

Let me have you look in your notebook to PTX-730.

Pazzani - direct
1

A.

PTX-730.

2

Q.

And what is that document?

3

A.

It's a Google document entitled Languages and

4

847

Personalized Search.

Yes.

5

MR. NELSON:

I will move to admit 730.

6

MR. VERHOEVEN:

7

THE COURT:

8

(PTX-730 is admitted into evidence.)

9

MR. NELSON:

No objection.

It's admitted.

Pull up slide 156, please.

10

BY MR. NELSON:

11

Q.

What does this document say?

12

A.

It discusses an experiment that was run at Google;

13

and essentially what it says is that the

14
15
16
17
18
19

MR. NELSON:

Let me pull up slide 157 which is

20

part of PTX-24.

21

BY MR. NELSON:

22

Q.

23

the estimating probability for the session category profile?

24

A.

25

profiler, and it uses the same algorithms as the rephil

What does this tell you about the algorithm that does

Yes, the session category is our fifth and final

Pazzani - direct

848

1

except it looks at the short term, the session-based profile

2

instead of the long term profile.

3

things you have just done in

4

Q.

5

category profile and the rephil profile?

6

A.

So that is looking at

And what is the estimated probability of the session

What I described earlier, it's

7
8
9

MR. NELSON:

And let me look at Bryan Horling's

testimony, the next slide, 158.

10

BY MR. NELSON:

11

Q.

What does Mr. Horling say about this?

12

A.

Well, how does the Kaltix Twiddler apply the session

13

category profile -- the session category user profile to the

14

group of documents?

15

And it says:

16
17
18
19
20
21

MR. NELSON:

22

BY MR. NELSON:

23

Q.

24

be?

25

A.

May I have the next slide, please.

And what did the Court define as unseen document to

It's a document the user has not previously seen.

849

Pazzani - direct
MR. NELSON:

1

Let me have slide 160, please.

2

BY MR. NELSON:

3

Q.

And how many web pages does Google currently have?

4

A.

According to a recent Google website, there are 60

5

trillion individual web pages.

6

Q.

7

documents -- the probabilities estimated for being unseen by

8

the user?

9

A.

10

And what does that tell you about how many

Well, hopefully there is a lot of unseen documents

since most people don't seen even one trillion or a billion.

11

MR. NELSON:

12

BY MR. NELSON:

13

Q.

14

Google's PTX-17?

15

A.

16

categories.

17

that it updates

Let me put up PTX-17.

And this is a portion of -- what is this figure from

It shows the number of documents of various
There are some documents like news websites
and others

18

And it shows sort of a

19

hierarchy of documents.

20

category.

21

news-related documents that are updated

22

There are

in one

And there is fewer documents that are saying

MR. NELSON:

Let's now turn to Search Ads and

23

talk about the estimating a probability that an unseen

24

document is of interest to a user.

25

slide 163.

And can you pull it up?

And let's go to PTX-115,

Pazzani - direct
1

BY MR. NELSON:

2

Q.

3

is here and how it is used?

4

A.

5

850

circled, and

And can you explain where the estimated probability

Yes.

So it happens in the SmartAds system that is

6
7
8
9
10

Q.

11

bottom?

12

A.

What portion does that?

That little genie guy on the

Well,

13
14
MR. NELSON:

15

And let me direct your attention,

16

let me pull up PTX-402, slide 164.

17

BY MR. NELSON:

18

Q.

So what is the SmartAds system?

19

A.

Well, it says that SmartAds is one of the most

20

important systems for maintaining and improving the quality

21

of ads that Google serves.

22

MR. NELSON:

Let me turn to slide 166.

23

BY MR. NELSON:

24

Q.

25

learning machine, user model, what portion of the SmartAds

Can you explain, the portion going back to our

851

Pazzani - direct
1

or what portion of the Search Ad system performs Element

2

1(e)?

3

A.

Yes.

So you may recall the

4
5
6
7
8
9
10

MR. NELSON:

Can you pull up PTX-401?

And the

11

pullout.

12

BY MR. NELSON:

13

Q.

14

estimating a probability?

15

A.

16

regression equation that I showed earlier, the log linear

17

model.

What does this portion of PTX-401 tell you about the

It uses an

.

That is that type of

18
19

So in some ways,

20
21
22
23

MR. NELSON:

24

BY MR. NELSON:

25

Q.

Let me have slide 168, PTX-402.

Can you explain

a little further?

852

Pazzani - direct
1

A.

This goes on to describe that logistic regression is

2

used, and it predicts the probability of the click-through

3

rate, and these have these 0s and 1s.

4

as I was drawling those equations, if

And I think you saw

5
6
7
8

MR. NELSON:

9

slide and just the pullout as well.

10

Q.

12

The

BY MR. NELSON:

11

Let me have PTX-397, please.

A.

And what is this?

13
14
15
16
17

Q.

And this element talks about predicting, estimating a

18

probability that an unseen document is of interest to the

19

user.

How is this doing that when it's looking at

21

A.

It's still calculating the

22

In this case, it's just a lower

23

Q.

Let me have you look in your notebook at Exhibit 942.

24

A.

942.

25

Q.

What is that?

20

Okay.

.
.

853

Pazzani - direct
1

A.

This is a Google document humorously titled The UBAQ

2

Virtuous Pumpkin.
MR. NELSON:

3
4

I move that 942 be admitted into

evidence.

5

MR. VERHOEVEN:

6

THE COURT:

7

(PTX-9432is admitted into evidence.)

8

MR. NELSON:

9

No objection.

It's admitted.

Can I have slide 170, with PTX-942

on it.

10

BY MR. NELSON:

11

Q.

What is Exhibit 942?

12

A.

942 describes some

13

Q.

What

14

A.

They are the

15

.

are described?

.

16

Q.

17

ad system?

18

A.

19

And how many ads are currently in the system, in the

I think I saw something about
MR. NELSON:

Let me put up PTX-400.

20

I think it's this graphic from PTX-400 says 20.

21

BY MR. NELSON:

22

Q.

23

shows the ads that are unseen by the user?

24

A.

The next.

25

So what leads you to conclude that Google's system

Well, the fact that hopefully you have not seen
You have seen a fewer number of that.

854

Pazzani - direct
MR. NELSON:

1

Let's turn to the Content Ads

2

element here.

The next slide please, PTX-222.

3

BY MR. NELSON:

4

Q.

5

estimating a probability of Content Ads?

6

A.

7

estimates the probability the user will click on a document,

8

and part of that equation is the probability the user is

9

interested in that document.

So let's talk about the Content Ads.

What is the

Again, there is a log linear regression equation that

In this case, the document is

10

an ad.

And that is based on the short term phil clusters.

11

So if you have recently clicked on Volkswagen ads, it will

12

show you more Volkswagen ads.

13

Q.

14

system?

15

A.

Is that also done by a portion of the SmartAds

16
17

MR. NELSON:

Let's go to slide 176.

Next slide.

18

176.

19

BY MR. NELSON:

20

Q.

21

estimated or defined by the user model in the Content Ad

22

system?

23

A.

24

parameters are the rephil clusters associated with

25

And so what is the learning machine having parameters

Well, here, the ads are the documents and the

855

Pazzani - direct
1
2
3
4
5
6
7
8

Q.

Let me just skip a little bit ahead here to PTX-413,

9

slide 179.

Actually, what is in your notebook?

10

PTX-413 first?

11

A.

What is

PTX-14 (sic) is a Google document called CUBAQ Update.

12
13

MR. NELSON:

I move that 413 be admitted into

evidence.

14

MR. VERHOEVEN:

No objection, Your Honor.

15

THE COURT:

16

(PTX-413 is admitted into evidence.)

17

MR. NELSON:

It's admitted.

Can you put up slide 179, please?

18

And the pullout.

19

BY MR. NELSON:

20

Q.

And what does this document tell you?

21

A.

This document talks about the CUBAQ profiler and how

22

it is used to estimate the probability a user is interested

23

in a document.

24
25

And essentially

856

Pazzani - direct
1
2

Q.

Let me have you turn to Exhibits 1468 -- or 1458 and

3

1462 in your notebook.

4

A.

1462.

5

Q.

I'll do them at the same time and just ask you to

6

tell me what they are.

7

A.

1458 and 1462.

8

Q.

Correct.

9

A.

They explain what certain numbers in the Google

10

system means.

These numbers are the numbers associated with

11

rephil categories, and they explain in English what these

12

numbers mean.

13

MR. NELSON:

14

Move that those two exhibits be

admitted into evidence.

15

MR. VERHOEVEN:

16

THE COURT:

17

MR. NELSON:

18

THE COURT:

19

No objection, Your Honor.

Okay.
Can I have slide 180.
You need me to say for the record

those are both admitted.

20

MR. NELSON:

Thanks.

21

(PTX-1458 and PTX-1462 are admitted into evidence.)

22

MR. NELSON:

Let me have the next slide.

23

BY MR. NELSON:

24

Q.

Can you explain what is on slide 180?

25

A.

Yes.

On slide 180, there is a number 24001.

I'm not

Pazzani - direct
1

857

sure if you can zoom in on that at the top (indicating).

2

And it says, that corresponds to the concept of

3

making money and earning money, and things of that sort.

4

Q.

Is that a short term phil cluster?

5

A.

Well, it's a phil cluster which could be in someone's

6

short term profile.

7
8

Q.

Let me have you turn to Exhibit 1457 in your

9

notebook.

10

A.

Okay.

11

Q.

And what is that document?

12

A.

1457 is a document that Google provided to PUM's

13

attorney which explains the coefficients associated with a

14

regression equation.

15

MR. NELSON:

16

And I would move that 1457 be

admitted into evidence.

17

MR. VERHOEVEN:

18

THE COURT:

19

(PTX-1457 is admitted into evidence.)

20

MR. NELSON:

21

Q.

It's admitted.

Let me have the next slide, 181.

BY MR. NELSON:

22

No objection, Your Honor.

Can you explain what this is?

23

Can you zoom this in?

24

And then, Dr. Pazzani, can you explain what is

25

being shown here?

858

Pazzani - direct
1

A.

Yes.

So what this is saying, the title, the top, at

2

the top, it was the ad phil cluster and the short term phil

3

cluster.
So what this is basically saying is if you have

4
5

this category in your short term profile and the ad is of

6

this category, then it's more likely that you will click on

7

the ad.

8

it's in your profile, you are less likely to click on an ad

9

of that category.

10

And then there is another ad category here that if

I know there are lots of numbers.

We wanted to

11

be able to show the names as they were on the previous

12

slide.

13

Q.

And is the .1469, is that the estimated probability?

14

A.

No, that is not the estimated probability.

15

would be a portion of the data used for it.

16

all the users.

That

That is true of

17
18
19
20
21

MR. NELSON:

Let me have the next slide, 182.

22

BY MR. NELSON:

23

Q.

24

ads, probabilities for ads that are unseen by the users?

25

A.

And does the Content Ad system also provide estimated

Yes, it does.

Again, there is

, some

Pazzani - direct
1

859

of which are unseen.

2

MR. NELSON:
I'm sorry.

Let me have slide 188.

3

Oops.

182 -- or 183.

I got ahead of myself.

4

BY MR. NELSON:

5

Q.

6

Search Ads and YouTube, Content Ads and YouTube practice

7

each of the aspects of Step 1(e)?

8

A.

9

probability that a document is of interest to a particular

And in your opinion, does each of Google Search,

Yes, they do.

10

user.

11

Each of them has an estimates the

And I'll read the rest but it's kind of small, but

each of them does.

12
13

MR. NELSON:

We should be able to move through

the rest of this hopefully more quickly.

14

So let's go to Element 1(f) of the '040 patent.

15

THE COURT:

16

I guess for the record, we now have

the proper chart up; correct?

17

MR. NELSON:

Correct, Your Honor.

18

THE COURT:

19

MR. VERHOEVEN:

20

MR. NELSON:

Okay.

Thank you.

Thank you, Your Honor.

So let me get the claim language

21

for, on slide 185.

22

BY MR. NELSON:

23

Q.

Can you read Element 1(f)?

24

A.

Yes.

25

automatic, personalized information services to the user.

Using the estimated probability to provide

860

Pazzani - direct
1

Q.

And is this done, for Search Ads, Content Ads and

2

Search, is this all done automatically?

3

A.

Yes.

4

Q.

And the same with YouTube?

5

A.

Yes.

6

Q.

And so let's look at slide 186.

7

animation?

8

A.

9

to the user.

Yes.

Can you explain that

So here are the possible documents of interest
And there is a user whose profile we know.

10

And this is showing that that user has selected a subset of

11

those documents based on the probability the user is

12

interested in those documents.

13

call -- that Google calls

14

MR. NELSON:

15

Q.

17

Let me have slide 187, PTX-44.

BY MR. NELSON:

16

That's the process that we

A.

What is the Kaltix Twiddler?

18
19
20

Q.

And let me have -- look in your notebook at PTX-39.

21

A.

Yes.

22

Q.

And what is PTX-39?

23

A.

PTX-39 is a declaration of Craig Sosin.

24

Q.

And who is Craig Sosin?

25

A.

He's a software engineer at Google.

Pazzani - direct
1

MR. NELSON:

2

861

And I move that Exhibit 39 be

admitted into evidence.

3

MR. VERHOEVEN:

4

THE COURT:

No objection, Your Honor.

It's admitted.

5

(PTX-39 was admitted into evidence.)

6

MR. NELSON:

Let me have slide 188.

7

BY MR. NELSON:

8

Q.

9

operation of the Kaltix twiddler?

And what does Mr. Sosin and Google say about the

10

A.

Okay.

He's describing in a little detail how it

11

works.

12

Kaltix twiddler does several things.

So for any particular web page search result, the

13
14
15
16
17
18
19
20
21
22
23

Q.

24

the link, dilip and category NavBoost aspects of this.

25

And let's go to the next slide, 189, and talk about

So what twiddler is responsible for applying the

Pazzani - direct
1

probability for these three, these three elements?

2

A.

3

862

the parts that are used

Well, the twiddler is called the Kaltix twiddler, but

4
5
6

Q.

Let me have you turn in your notebook -- actually,

7

let me put up slide 190, please, and pull out, this is

8

PTX-200.

9

A.

And what is this code?

This is part of the Kaltix twiddler, and this

10

actually describes how it works, and it's -- what it's

11

basically saying is, for a --

12
13
14
15

So we ask what would happen if

16
17
18
19
20
21
22
23
24

Q.

And let's talk next about, let me have slide 191.

25

Let's talk about the rephil and session categories, slide

863

Pazzani - direct
1

192, please.
And what does Mr. Sosin say about this?

2
3

A.

I'm sorry.

4

Q.

Yes.

5

had before.

6

aspect works the same?

7

A.

8

profiles as well as link and dilip profiles.

Yes.

Could you repeat the question?

I just asked, this is a repeat of the slide we
I guess really I'm just asking if the twiddling

The twiddling aspect works the same for rephil

9
10
11
12

Q.

And let me have slide 193, please.

13

And what does Mr. Horling say about the

14

operation of the Kaltix twiddler?

15

A.

16
17
18
19
20
21

Q.

And could I have slide 194, please.

22

walk through now by each element how Google's search

23

practices claim 1 of the '040 patent?

24

A.

25

That's like recording the queries and the search result

Yes.

Can you just

So first we saw transparently monitoring.

864

Pazzani - direct
1

clicks that you clicked on.

Then there's updating the

2

user-specific data files.

3

associated with an individual user associated with their ID.

That's storing in Kansas the data

4
5

Then you estimate the parameters of the

6

learning machine.

7

categories of the documents that are in the user-specific

8

data files.

9

That's the user interest in the

In order to figure out those categories, you

10

have to have analyzed them.

And then you can estimate the

11

probability based on the user model that the user is

12

interested in a particular document.

13

for instance, to reorder the search results presented to the

14

user, so that different orders for different users based on

15

their profile, which is based on

And then you use that,

16
17

Q.

And let's turn now to Search Ads and Content Ads

18

together.

19

And this is a response to a request for admission.

20

does this say about how search and Content Ads are a high

21

probability?

22

A.

23

click-through rate as adjusted by UBAQ is used in the

24

auction that determines the ranking of candidate ads in

25

Google's AdWords system.

So let's talk about, let me have PTX-403, please.

Okay.

What

Google admits that the probability of the

865

Pazzani - direct
1

Q.

And let me have you refer in your notebook to

2

PTX-110.

3

A.

4

Google document.

PTX-110 is a document entitled Google AdRank.

5
6

And what is that document?

MR. NELSON:

I move that Exhibit 110 be admitted

into evidence.

7

MR. VERHOEVEN:

8

THE COURT:

9

It's a

No objection, your Honor.

It's admitted.

(PTX-110 was admitted into evidence.)

10

MR. NELSON:

May I have 197 on the board,

11

please?

12

BY MR. NELSON:

13

Q.

This is PTX-110, so what does this tell you?

14

A.

This talks about the auction that decides which ads

15

are shown to the user and ads are displayed in the order of

16

, highest at the top, lowest at the bottom.

And

17
18
19

Q.

And how is the estimated probability that's

20

associated with the step 1(e) used in Search Ads?

21

A.

22

more likely to click on are more towards the top, but this

23

can be affected by the bid as well.

24

Q.

25

about Search Ads, the ads that the user isn't interested are

Well, all things being equal, those the users are

In this case, it's the opposite because we're talking

866

Pazzani - direct
1

less likely to be shown; is that right?

2

A.

3

ads on the bottom, for instance.

4

Q.

5

Mr. Gopalratnam said about this?

6

A.

Yes.

12

Q.

And how is this done in Content Ads?

13

A.

In Content Ads, it's based on

15

Q.

And let me have PTX -- slide 199.

16

the calculation used?

17

A.

18

way of computing the click-through rate based on the user

19

model.

20

Q.

Yes.

So the Amazon ads might be on top and the eBay

Let me have slide 198, please.

Can you tell me what

When asked how the ad rate is termed, he said

7
8
9
10
11

14

Yes.

Uses the same auction.

And is this also

It's just a different

And let me have slide 200.

21

And can you summarize your opinions

22

regarding whether Element 1(f) is practiced by Search,

23

Search Ads, Content Ads and YouTube?

24

A.

25

used to provide automatic personalization to the user.

Yes.

To estimate the probabilities in each case RE

867

Pazzani - direct
1

Q.

2

slide 201, please.

3

A.

Yes, it is.

4

Q.

You have already given your opinion that -- well,

5

tell the jury what a dependent claim is.

6

A.

7

this case, it depends on claim 1, so it has to do everything

8

that is in claim 1, the preamble steps A, B, C, D, E, but

9

also has to do this, where the monitored user interactions

10

And let's go on to dependent claim 22.

Can I get

Is this dependent claim 22?

So a dependent claim depends on another claim.

In

include a sequence of interaction times.

11

So all we have to talk about this here, because

12

we've discussed this previously, is does Google use the

13

sequence of interaction times.

14

Q.

15

this is part of Exhibit PTX-1268.

16

going on here?

17

A.

18

for

19

part?

20

Let me have slide 204, please, the next one.

Yes.

And

Can you explain what's

This is part of a web history.

I think this is

Could I ask you to zoom in on this

Okay.

21
22
23
24
25

Q.

And let me have slide 205.

And can you pull up --

Pazzani - direct
1

and what is, what is this?

2

you explain what's shown here?

3

A.

4

keeps track of

Yes.

This is part of PTX-373.

868
Can

So this is a further example of how Google

5
6
7
8
9
10
11

Q.

And let me have you turn in your notebook to PTX-35.

12

And what is PTX-35?

13

A.

14

responses to the plaintiff's amended notice of Rule

15

30(b)(6), deposition of Google.

It's Google's supplemental responses, objections and

16

MR. NELSON:

17

I move that Exhibit 20 -- PTX-35 be

admitted into evidence.

18

MR. VERHOEVEN:

19

THE COURT:

20

No objection, Your Honor.

It's admitted.

(PTX- 35 was admitted into evidence.)

21

MR. NELSON:

So can you put that up?

22

BY MR. NELSON:

23

Q.

And what is being shown here?

24

A.

What is being shown here is explanation, discussing a

25

little bit more the details of that, what we just saw.

869

Pazzani - direct
So, for instance, if we can zoom in,

1

.

2
3

what this number means (indicating), and that number is the

4

.

5

computer way of saying

6

Q.

7

explaining

8

A.

, in essence.

?

That's correct.
MR. VERHOEVEN:

10

MR. NELSON:

11

THE COURT:

12

MR. NELSON:

13

THE COURT:

Objection, leading.

Can I continue?
Could you hear that?
Yes.
I wasn't sure.

Is that an

objection?

15
16

So that's a

And this is a statement by Google essentially

9

14

And we've asked

MR. VERHOEVEN:

I objected leading, but the

answer came out.

17

THE COURT:

The objection is overruled because

18

it's moot, but it was leading.

19

leading.

20

MR. NELSON:

Let's refrain from the

Okay.

21

BY MR. NELSON:

22

Q.

23

will ask you, is that a Google document?

24

A.

25

Let me have you turn in your notebook to PTX-378.

Yes.

It's a Google document.
MR. NELSON:

I move that PTX-378 be admitted.

I

870

Pazzani - direct
MR. VERHOEVEN:

1

No objection, Your Honor, with

2

the caveat that my copy is partially legible.

3

a legible copy.
THE COURT:

4

All right.

We'd request

Provide the better

5

copy if you haven't already.

6

(PTX-378 was admitted into evidence.)

7

MR. NELSON:

8

Q.

Pull up the pullup, please.

BY MR. NELSON:

9

Otherwise, it's admitted.

10
11

And so what does this slide tell you about how the
are used?

A.

12
13

, and I'm sure you don't remember, but
about an hour ago I discussed how the rephil profile worked

14
15
16
17
18
19
20
21
22
23
24
25

So what's that really saying in English rather
than mathematics,

871

Pazzani - direct
1
2
3
4
5

Q.

6

Let's talk about Search Ads now.

Can I get slide

209?
And can you tell me what this document is?

7
8

A.

Yes.

This is part of a web history, and it shows

9

something called the sponsored links, which is over there

10

(indicating).

11

clicked on in Google so it can display them to you.

12

Q.

13

1268, I should say.

14

can you tell me what's on PTX-373, page 1143717 with respect

15

to timestamps?

16

A.

And the sponsored links are the ads that you

This is also part of PTX-278.

Let me have -- or

Let me have slide 210.

And is this --

17
18
19

Q.

And how are these used in Google Search Ads?

20

A.

Well, one way is that if you -- one thing that it can

21

do is

22

and that is important to something Google calls a

23
24

Q.

And let's turn to, let's turn to Content Ads, slide

25

212, please.

872

Pazzani - direct
And can you speak about, this is Google's

1
2

response to our request for admission.

What does this tell

3

you with respect to

4

A.

Well, it says that Google predicts whether

7

Q.

And let me show you slide 213, PTX-404.

8

A.

So why that's useful is sometimes

Q.

Thank you, Dr. Pazzani.

?

5
6

9
10
11
12

Let me show you what is on PTX-404 and ask

13

what this tells you about the use of

in Content

14

Ads.

15

A.

Yes.

Q.

And let me have PTX-406, please and the pull out.

So this talks about how it computes

16
17
18
19
20
21
22

And what is this -- this talks about the

23

.

What does this say?

24

A.

This is the

the Google Content Ads.

25

saw this document before and essentially it

We

873

Pazzani - direct
1
2
3
4

Q.

Can I have the next slide, please, slide 215.
Can you summarize your opinion regarding

5
6

whether Search, Search Ads, Content Ads and YouTube infringe

7

claim 22?

8

A.

9

storing

Yes.

By the use of storing

in Kansas or

associated with user data, then you can derive

10

a sequence of

11

Q.

12

the '276 patent.

13

A.

Okay.

14

Q.

And this should go quite a bit quicker.

Okay.

Let's turn now to the other patent at issue,

Let me have slide 219, please.

15

And can you tell

16

me what is shown on this slide?

17

A.

18

patent.

19

there are also some parts that are different.

20

Q.

21

the preamble.

22

the difference in the preamble?

23

"automatic."

24

A.

25

one difference.

Yes.

This is comparing the '040 patent and the '276

There are many similarities between the two, but

And take a look at the non-colored portion first,

Yes.

What are the differences in the preamble or
Focus on the word

So the word "automatic" is not there, so that's

874

Pazzani - direct
1

Q.

Do you have, is your opinion with respect to

2

infringement the preamble of the claim 1 of the '276 the

3

same as for the '040?

4

A.

5

automatic, but if things were automatic, it still infringes.

6

Q.

And otherwise your opinion is the same?

7

A.

Yes.

8

Q.

And so let's look at the next, the next slide here,

9

which let's go to -- let's go to 223, actually.

Well, '276 is more general.

10

It does not require

So here we talked about users a lot in the

11

preamble.

12

A.

13

earlier for the preamble of claim 1 of the '040 patent and

14

all this is saying is, the evidence there is exactly the

15

same as the evidence here.

16

again, but the Gaia ID, the

17

cookie and the PrefID are used in exactly the same way and

18

infringe in exactly the same way.

19

Q.

20

0365, 407 and 406?

21

A.

I think you left 1312 out, but, yes.

22

Q.

Can you go to the next slide?

23

What is this slide?

This slide just summarizes the evidence that we used

We don't need to go over it
the DoubleClick

So relying in part on PTX-576, 1312, 0113, 1312,

And can you summarize your opinions whether

24

Google Search, Search Ads, Content Ads and YouTube practice

25

the preamble of the '276 patent?

875

Pazzani - direct
1

A.

Yes, they do.

2

MR. NELSON:

Let's turn to the next element, the

3

transparently monitoring element.

Can I have the next

4

slide, 226.

5

BY MR. NELSON:

6

Q.

7

says "normal use of a browser program" instead of "normal

8

use of a computer."

9

A.

Those are the two elements together.

Now, the '276

Is there any difference there?

Well, there is a difference but not a significant

10

difference with respect to the infringement because

11

everything we discussed earlier was happening within a web

12

browser.

13

ads, et cetera.

The user accessing the Google website, clicking on

14

MR. NELSON:

Let me go to slide 227.

15

BY MR. NELSON:

16

Q.

17

in -- summarize the evidence you relied on in your testimony

18

for the '040 patent.

And does this slide summarize -- let me just read it

19

Let me start over.

What is on this slide?

20

A.

This slide shows the evidence used on claim 1(a) of

21

the '040 patent.

22

1(a) of the '276 patent.

23

deposition, 395 for the Google Gopalratnam deposition,

24

PTX-0404 of the Ponnekanti deposition and the Zamir

25

deposition.

And that evidence can be used for claim
So Exhibits 11, 370 of the Horling

Pazzani - direct
1

MR. NELSON:

876

Let me have slide 228.

2

BY MR. NELSON:

3

Q.

4

Content Ads, and YouTube Content Ads, practice Element 1(a)

5

of the '276 patent?

6

A.

In your opinion, does Google Search, Search Ads,

Yes, for the same reasons I cited previously.

7

MR. NELSON:

8

slide.

9

Let's go to Element 1(b), the next

BY MR. NELSON:

10

Q.

And so what is the difference here between the

11

Element 1(b) of the '040 and the element here doesn't have

12

the letters on the '276.

13

A.

Sure.

14

Q.

The analyzing element.

15

A.

Yes.

16

documents of interest to the user.

17

complicated the way we monitor the data, update the user

18

specific data.

19

So one difference is we're determining the

MR. NELSON:

And it's less

Let me have slide 231, please.

20

BY MR. NELSON:

21

Q.

22

interest to the user."

23

A.

24

documents (electronic files, including text or any other

25

type of media) for which the user has a positive response."

Can you read the Court's definition of "documents of

Yes.

"Documents of interest to the user are

877

Pazzani - direct
MR. NELSON:

1

Let me turn to slide 233.

This is

2

part of the patent, PTX-1.

3

BY MR. NELSON:

4

Q.

5

in the patent?

6

A.

7

creates positive and negative patterns.

8

of documents of interest to a user:

9

are visited following a search query is an example of

10

Can you tell me sort of what a positive response is

Well, positive response -- I'm sorry.

The user

Positive examples

Search results that

expressing interest in something.

11

MR. NELSON:

And can we go to slide 234, please?

12

I'll skip 233 for now.

13

BY MR. NELSON:

14

Q.

And can you just explain what is shown on this slide?

15

A.

Yes.

This is going back to some of the

16

earlier.

And the URLs,

17
18
19

MR. NELSON:

20

Q.

22

Slide 235.

BY MR. NELSON:

21

And let me go to 235.

documents?

23

What does Mr. Horling say about analyzing the

Let me ask it this way:

24

Mr. Horling said about this element?

25

A.

Well, I think he is getting at

Can you tell me what

Pazzani - direct

878

1
2

the search results for the user in the future because Google

3

assumes you have expressed interest in the document.
MR. NELSON:

4
5

And let's go to the Search Ads

portion here.
Let me actually ask you, let's go back to slide

6
7

234 for a minute.

8

Let's go to 237.

9

Never mind.

BY MR. NELSON:

10

Q.

11

shown on this document?

12

A.

Let's go to Search Ads.

And this is PTX-375 again.

Yes.

Can you tell me what is

These are the document IDs associated with the

13
14
15

(Counsel confer.)

16

MR. NELSON:

Sorry about that.

17

BY MR. NELSON:

18

Q.

19

data to determine the documents that are of interest to the

20

user?

21

A.

22

clicks, it's determining which documents are of interest to

23

the user.

24

Q.

25

asked you that question for Google Search before so I'm just

So can you tell me, does Google analyze the monitored

Yes.

By monitoring the user's ad clicks or result

Is that same true for Search?

I don't know if I

Pazzani - direct
1

re-asking it.

2

A.

3

879

interest on the documents that you have clicked on.

Okay.

4

So, yes, the result clicks are a sign of

MR. NELSON:

5

please.

6

239, please.

7

BY MR. NELSON:

8

Q.

9

answered for us.

And let's turn to the next slide

Let's turn to Content Ads.

And may I have slide

And this is another Request For Admission that Google
Can you tell me what this says about

10

analyzing the monitored data to determine documents of

11

interest to the user?

12

A.

13

content-based system.

14

DoubleClick cookie

Yes.

And this is getting back to CUBAQ, the
Google stores data associated with a

15
16

Q.

And does Google analyze that data to determine

17

documents of interest?

18

A.

Yes, it does.

19

In particular,

MR. NELSON:

And let me turn to slide 240,

20

Mr. Zamir's testimony.

21

BY MR. NELSON:

22

Q.

And can you summarize what he said?

23

A.

Yes.

24
25

He discussed again

880

Pazzani - direct
MR. NELSON:

1

Can I have slide 241, please?

2

BY MR. NELSON:

3

Q.

4

say about determining user interest in ads and analyzing

5

data?

6

A.

7

interested in, he said:

And this is Mr. Weinberg from yesterday.

What did he

When asked how does Google know what a cookie is
So I --

8
9
MR. NELSON:

10

And can I have slide 242, please.

11

BY MR. NELSON:

12

Q.

13

about the analyzing the monitored data to determine

14

documents of interest for the user?

15

A.

16

the ID in the cookie, and that's keeping track of the ads

17

the users has clicked on.

This is Ms. Illowsky's testimony.

What did she say

So she discussed that the ad ID is associated with

18

MR. NELSON:

Can I have the next slide, please?

19

BY MR. NELSON:

20

Q.

21

Google Search, Search Ads, Content Ads and YouTube practice

22

Element 1(b) of the '276 patent?

23

A.

1(c)?

24

Q.

I think it's 1(b).

25

A.

Oh, I'm sorry.

And can you summarize your opinion whether or not

The preamble is 0.

Yes.

So by storing the data in

881

Pazzani - direct
1

Kansas and then using this to assume the user's interest in

2

those things the user has clicked on, generating profiles

3

based on those clicks.

4

the data to determine documents of interest.

5

MR. NELSON:

That shows that Google is monitoring

Okay.

Let's turn to the next

6

element, 1(c).

7

BY MR. NELSON:

8

Q.

9

specific learning machine based at least in part of the

10

The next claim.

So this claim says:

Estimating parameters of a user

documents of interest to the user.

11

Do you see that?

12

A.

Yes.

13

Q.

And this element is fairly similar to the last one.

14

Can you explain what is different?

15

A.

16

talking about a user specific learning machine.

Well, it doesn't mention the user model.

17

MR. NELSON:

It's just

Let's go to slide 246 real quick.

18

BY MR. NELSON:

19

Q.

20

machine."

The Court also construed "user-specific learning

21

machine."

Can you read that definition to the jury?

22

A.

23

machine (as construed) specific to the user."

24

Q.

25

well, let me ask it differently.

So this is the Court's definition of "learning

Yes.

"A user-specific learning machine is a learning

Is the learning machine made user specific in '276 --

Pazzani - direct
1

882

Is the user-specific learning machine in '276

2

user specific?

3

A.

4

user's data such as the data stored in Kansas.

5

Q.

The same as for the '040?

6

A.

That's correct.

7

Content Ads, Content Ads in YouTube and Search, the argument

8

is the same.

9

Q.

10

Yes, it's user specific because it operates on the

So for each of them, Search Ads,

For each of the different user -- Well, let me go to

slide 250, please.

11

Can you identify what the user-specific learning

12

machine is for this element for each of the five accused

13

Google Search properties?

14

A.

15

link profiler, the Kaltix Twiddler.

Yes.

16
17

The

Or,

A portion of the Kaltix Twiddler plus the user's
link profile.

18
19

They're the things we discussed earlier.

The dilip profiler, portion of the Kaltix
Twiddler, and the user's dilip profile.

20

Similarly, the rephil and the user's rephil.

21

And the NavBoost, et cetera.

22

Q.

And this element talks about determining or

23

estimating the parameters of the user-specific learning

24

machine based at least in part on the documents of interest

25

to the user.

Is that element met in your opinion?

883

Pazzani - direct
1

A.

Yes, it is.

2

Q.

And is the way that it's met, the evidence you are

3

relying on the same as you relied on for the '040 patent?

4

A.

Yes, the evidence is the same.

5

MR. NELSON:

Let me have slide 251, please.

6

BY MR. NELSON:

7

Q.

8

evidence you are relying on for this element?

9

A.

10

Can you just read into the record the summary of the

Yes.

and 0376.

11
12

So there is each profiler, which is in PTX-0770

And then for link, it's PTX-30, 33, 34, 76 and
Glen Jeh's testimony.

13

And then for dilip, it's 0025, 69, 98, 379.

14

And for Category NavBoost, it's 0030 and 0038.

15

Q.

Is the data in the

, the document identified

16

as in the

17

is used to estimate the parameters of the user specific

18

learning machine?

, is that part of the information that

19

MR. VERHOEVEN:

20

THE COURT:

Objection, leading.

Sustained.

21

BY MR. NELSON:

22

Q.

23

the portion of Kansas data that is used to estimate the

24

parameters of the user specific learning machine?

25

A.

Can you identify the parameters?

Yes.

It's the

Can you identify

Pazzani - direct

884

1
MR. NELSON:

2

And let me turn to slide 252.

3

BY MR. NELSON:

4

Q.

5

well, what evidence did you rely on for your opinion for the

6

rephil and session category profiles as whether they

7

practice this -- whether Google Search rephil and Search

8

session category practice this element?

9

A.

10

Can you summarize your opinion for this element --

So if you recall, these use the

and

it's PTX-0030, 34, 37, 69.

11

PTX-0030 and PTX-0069 for session category.

12

Q.

13

estimate those parameters?

14

A.

15

And what column in Kansas is at least in part used to

Again, it's the
MR. NELSON:

.
Let's talk, let's go to Search Ads.

16

And can I get slide 254, please.

17

Content Ads together.

18

BY MR. NELSON:

19

Q.

20

relied on for Search Ads and Content Ads meeting claim 1,

21

element C of the '040 patent, the same evidence you are

22

relying on for meeting this element of the '276 patent?

Can you summarize the evidence -- is the evidence you

23

MR. VERHOEVEN:

24

THE COURT:

25

We'll do Search Ads and

BY THE WITNESS:

Objection, leading.

Overruled.

You can answer.

Pazzani - direct
1

A.

Yes.

2

885

parameters of the learning machine.
In Search Ads, it's PTX-112, 113, 397, 398, and

3
4

The same evidence was used for estimating

869.

5

And in Content Ads, it's 223, 404 and the Zamir

6

deposition.

7

Q.

8

learning machines for both Search Ads and Content Ads?

9

A.

10

Can you identify the respective user-specific

Well, for Search Ads, the user-specific learning

machine is the UBAQ profiler

11
12

Q.

What about for Content Ads?

13

A.

In Content Ads, that profiler wasn't given a name but

14

it's the thing that

15
16
17

MR. NELSON:

Let me turn to the next slide,

18

please.

19

BY MR. NELSON:

20

Q.

21

Element 1(c) of the '276 patent meets each of the accused

22

products?

Does this slide 255 summarize your opinions whether

23

MR. VERHOEVEN:

24

THE COURT:

25

BY MR. NELSON:

Objection, leading.

Overruled.

886

Pazzani - direct
1

Q.

Go ahead.

2

A.

Yes, it does.

3

Q.

That's what the green checkmarks means on all these

4

charts?

5

A.

That's correct.

6
7

THE COURT:

Let me see counsel at sidebar,

please.

8

(Sidebar conference held.)

9

THE COURT:

10

It's just past 4:00 o'clock.

Are

you expecting to finish your direct before 4:30?

11

MR. NELSON:

12

chance we will yet.

13

I think there is a pretty good

quickly a possible.

14

I'm just trying to get through this as

MR. VERHOEVEN:

Just so you know, I forgot what

15

slide you left off on, but we're here (indicating slides

16

remaining in the binder).

17

MR. NELSON:

Was it one that was not fixed?

18

MR. VERHOEVEN:

No, no, no.

I'm not complaining

19

about that, but the question.

20

257, and the last slide of the deck is 344.

21
22

THE COURT:

I believe we're around slide

Are there still approximately 100

slides left?

23

MR. NELSON:

24

very fast.

25

have to go through.

Yes, there are.

A lot of these go

There are a couple elements coming up that we

Pazzani - direct
1

THE COURT:

2

MR. NELSON:

3

THE COURT:

887

Okay.
They go very fast.
I thought we should talk.

In the

4

event that you are likely to finish around 4:20, 4:25 I was

5

curious what Mr. Verhoeven might want to do but it seems

6

like it's unlikely that you are going to finish even with a

7

few minutes left.

8
9

MR. VERHOEVEN:

Your Honor, if we only have

five minutes left, I don't think that's enough time for me

10

to finish my first cross module.

11

left, I would request the ability to begin cross.

12

THE COURT:

13

15 minutes left.

14

Okay.

But if we have 15 minutes

So certainly if there is

would your request be?

15

If there is 5 to 15 minutes left, what

MR. VERHOEVEN:

Well, I'd like the ability to,

16

if I start a module, finish it.

17

minimum I would need to do that.

18

THE COURT:

19

MR. VERHOEVEN:

20
21
22

any time you want.

I think 15 is probably the

Okay.
But I would be happy to start

Let me know and I'll start.

THE COURT:

You can't start until he is done

with his direct.

23

MR. VERHOEVEN:

24

MR. FRIEDMAN:

25

MR. VERHOEVEN:

Right.
You can't start until he says so.
Whatever Your Honor wishes I

888

Pazzani - direct
1

will do.

2

My first module is 15 minutes.
THE COURT:

I'm not going to make you start with

3

less than 15 minutes unless you tell me that is what you

4

want me to do, and I am hearing you don't want to do that.

5

MR. VERHOEVEN:

I think ...

6

(Counsel confer.)

7

MR. VERHOEVEN:

Yes.

We'll just -- okay.

8

Mr. Horwitz, go ahead and tell them.

9

MR. HORWITZ:

10

Your Honor, two things.

If it's

less than 15 minutes, we don't want to have to eat time.

11

THE COURT:

I'm not talking about eating the

12

time.

13

early and nobody is charged for the time.

14
15
16
17

I'm talking about letting the jury go a little bit

MR. HORWITZ:

Second, we don't want them to be

able to talk to the witness overnight.
THE COURT:

Right.

You would want it to be

essentially the cross has begun and they can't talk.

18

MR. HORWITZ:

19

THE COURT:

Exactly.

I would agree to that as well in

20

light of the circumstances, but this is all, just so we're

21

on the same page, if we finish by 4:15 or earlier with the

22

direct, we'll just go right into cross, but I'll still cut

23

you off about 4:30.

24

I'm going to cut you off at 4:30, but we won't start the

25

cross until the morning.

If we finish with direct after 4:15,

Okay?

889

Pazzani - direct
1

MR. NELSON:

That's fine.

2

MS. JACOBS:

Your Honor, given that Dr. Pazzani

3

has been on the stand all day, is it really fair to him to

4

start cross 15 minutes after six and-a-half hours?

5

THE COURT:

If we're done by 4:15 with the

6

direct, we're going to start the cross, fair or not.

7

don't think, it doesn't seem very likely but I'm not going

8

to hold them to it.

9

MR. NELSON:

10
11

THE COURT:
further.

But I

15 minutes is not going to happen.
I'm not going to slow you down any

Let's see what happens.

Okay.

12

(Sidebar conference ends.)

13

THE COURT:

14

MR. NELSON:

You may continue.
All right.

15

please?

16

BY MR. NELSON:

17

Q.

18

Can I get slide 257,

wasn't in the '040 patent.

Let's talk about Element 1(d).

19
20

Next slide, please.

And this element

Well, this is

receiving a search query.

21

Could I get slide 259, please.

And this is

22

a picture of the Google Search screen.

23

search queries?

24

A.

25

box, clicks on search Google and Google receives the search

Yes.

Does Google receive

The user types search queries into the search

890

Pazzani - direct
1

query.

2

Q.

3

PTX-22, that figure?

4

A.

Yes.

5

Q.

And let me also put up PTX -- well, PTX-17 real

6

quick.

7

A.

8

It says what we're talking about.

9

query.

And let me put up slide 260, and that's shown in

What does that say?
This is a Google document entitled life of the query.
The system receives the

10

Q.

And let's talk about search address.

11

slide 262, which is a portion of 115.

12

document show?

13

A.

14

from the user.

15

web server, et cetera.

16

Q.

17

Can you put up

Content Ad system.

And what does this

It shows that the system receives the search query
It goes to the Google front end, the Google

And let me put up slide 264 and let's talk about the

18

What is this document?

What are search referral

19

terms?

20

A.

21

The user doesn't explicitly type a query into the Google

22

Search box, but if the Google types a query into Bing and

23

goes to the Los Angeles Times, what happens is there's a

24

refer in the HTML and the http, and that refer indicates

25

what query one came from previously, or what website one

Yes.

On content ads, it's actually much more subtle.

891

Pazzani - direct
1

came from previously.

2

So a Bing search result page will contain

3

what's called the search referred term.

It's the queries

4

typed to Bing or the queries typed to Google.

5

And if you go to the L.A. Times, the search

6

referred terms from the search engine that got you there are

7

a part of these search referred terms that are then passed

8

to the Content Ad system.

9

Q.

And let me have you look in your notebook for

10

Exhibit 416.

11

A.

Okay.

12

Q.

And tell me what that document is.

13

A.

416 is a Google e-mail from Oren Zamir.

14
15

MR. NELSON:
evidence.

16
17

MR. VERHOEVEN:

One minute, Your Honor.

No

objection.

18
19

I ask that that be admitted into

THE COURT:

It's admitted.

(PTX-416 was admitted into evidence.)

20

MR. NELSON:

Can you put up Exhibit 416?

21

BY MR. NELSON:

22

Q.

23

and search queries?

24

A.

25

Google, Google queries as well as those from Yahoo, et

What does this document indicate about Content Ads

Actually, that describes just what I did, that

892

Pazzani - direct
1

cetera, used today, they're added with high weights to the

2

documents they lead to.

3

Queries from non-Google Search engines are

4

also associated with the DoubleClick cookie and used within

5
6
7

Q.

What does RPM stand for?

8

A.

I think it's revenue per million.

9

Q.

Let me put up slide 266, please.

What did Ms.

10

Illowsky say about referred terms?

And just look at the

11

bottom of the slide.

12

A.

13

and there's a referral URL that says what page it came from.

14

And so some search engines choose to include the query in

15

that, and so the term that was used as the query is in that

16

URL.

17

Q.

So she described what the search referred terms are,

And let me turn to the next slide, please.

18

And can you tell me what this, can you

19

summarize your opinions with respect to whether the accused

20

products practice this element?

21

A.

22

the use of the search box and Google Content Ads does

23

through the use of, use of the search referral terms.

24

Q.

25

retrieving a plurality of documents based on the search

Yes.

So Google Search does and Search Ads through

And let's talk about the next element.

This is

893

Pazzani - direct
1

query.
Do you see that?

2
3

A.

Yes, I do.

4

Q.

And this one is also not in the '040 patent, so let's

5

talk about this one briefly.

6

MR. NELSON:

Can I have slide 271, please?

This

7

is PTX-17.

8

BY MR. NELSON:

9

Q.

What does this document say?

10

A.

Well, after the system receives the query, it finds

11

matching results, and that's retrieving documents based on

12

the query and then it presents them to the user.

13

there's a little more that happens between those two steps.

14

Q.

15

does this document say?

16

A.

And let's put up slide 272.

I think

Go ahead and -- what

This is describing the Kaltix twiddler and it says,

17
18
19

Again, it's discussing these

20

candidate search results where the retrieval of documents

21

based on the query.

22

Q.

23

get PTX-115, slide 274.

24

PTX-115 show that documents are retrieved?

25

A.

And this is PTX-24.

Yes, it does.

Let's go to Search Ads.

Can I

And does this document show, does

We have not discussed it a lot, but

894

Pazzani - direct
1
2
3

Q.

And let's put up the next slide, 276, and talk about

4

Content Ads.

5

what does this tell you about documents being retrieved in

6

response to a query?

7

A.

8

added with high weight to the documents they lead to.

9

let me see if I can explain that.

And can you tell me, we just had this one up,

Well, it talks a little bit about how the query is
So

When you have a page on

10

the Los Angeles Times, like about a Chevy Volt, if you got

11

there by typing hybrid Chevy Volt or electric Chevy Volts,

12

or just electric car, then it's as though that page had

13

those search queries, the search query in it.

14

So the word "electric" would be added to

15

the Los Angeles Times page.

And then it would be Googled

16

Content Ads, would retrieve ads related, for instance, to

17

electric cars.

18

Q.

19

what Ms. Illowsky says regarding this topic?

20

A.

21

that out.

22

found in the page.

23

in a page?

24

page.

25

the page.

And let me turn to slide 277.

Yes.

And can you tell me

So he describes what I just did.

We find the query.

So we parse

And then we pretend it was

What do you mean by pretend it was found

It becomes one of the textual signals from the

Those are the signals used to retrieve ads related to

895

Pazzani - direct
MR. NELSON:

1

And could I have slide 278, please?

2

BY MR. NELSON:

3

Q.

4

patent, whether each of the accused Google products

5

practices the retrieving of plurality of documents step in

6

the '276 patent?

7

A.

8

user, it then retrieves documents related, or based on that

9

search query, and Google Search and Search Ads obviously do

And can you summarize your opinion whether the '276

Yes.

So having received the search query from the

10

it in a more subtle way than Google Content Ads do.

11

Q.

12

explain what the color coding means on this slide?

13

A.

14

patent.

15

are combined to the claim of the '276 patent that's

16

highlighted.

17

Q.

18

retrieved documents, is the evidence you relied on the same

19

as you relied on for the analyzing documents step in the

20

'040 patent?

21

A.

Yes.

24

Q.

And let me turn to slide 282.

25

into the record the evidence that you relied on that this

Can I have the next slide, 280.

Yes.

And so can you

So what we see is claims D and E of the '040

And in many ways, claims D and E of the '040 patent

And with respect to the identifying properties of the

Though these are things like the

22
23
And can you just read

896

Pazzani - direct
1

element was practiced in the '276 patent as well?

2

A.

3

PTX-0025, PTX-00 -- I'm sorry.

4

Horling's testimony.

Yes.

5
6

For the link categories, it's PX-0024,
0443, 037, and Brian

For the dilip categories, it's PTX-0025,
PTX-0376, and Haveliwala's testimony.

7

And for category NavBoost, how it uses the link

8

and dilip categories, it's PTX-0030.

9

Q.

And let's turn to slide 284, please, skip this one.

10

And can you summarize the evidence you relied on

11

with respect to the rephil and Session Category for the

12

identifying -- for identifying the property of the retrieve

13

document aspect of this element of '276?

14

A.

15

category system and that's described in PTX-0024, PTX-0025,

16

and PTX-0030.

17

Q.

18

some more evidence with respect to the identifying document

19

properties.

Can you just read that into the record?

20

A.

So rephil and Session Category use PTX-0024,

21

PTX-0876, and PTX-0027.

22

Q.

23

about the applying the identified properties of the of the

24

retrieved documents to the user specific learning machine to

25

estimate a probability stat?

Yes.

So both of these profilers use the rephil

And let's go to the next -- slide 286.

Yes.

And the next slide, please.

And this is

And can we now talk

897

Pazzani - direct
Could I have the next slide, please?

1

And

2

remember, before we were talking about rank by long click

3

probability and these three profiles.
What does this slide represent?

4
5

A.

6
7
8

Q.

And can you read the evidence that you relied on into

9

the record?

10

A.

Yes, I can.

PTX-0024, PTX-0097, PTX-0200, Brian

11

Horling's testimony, Glen Jeh's testimony and the Jahr

12

testimony.

13

Q.

14

probability, is your opinion for the '276 estimation any

15

different than your opinion for the '040 step?

16

A.

No, it's not.

17

Q.

And let's talk about -- let's go to the next slide,

18

290.

19

Category.

PTX-0382, PTX-0385, PTX-0433, and PTX-0729.

And is your opinion, regarding the estimation of

It's the same.

And let's talk about the rephil and the Session

20

What is your opinion whether there is the,

21

applying the identified properties of the retrieved

22

documents to the user specific learning machine to estimate

23

a probability step is met, or a portion of this step is met

24

for these two profiles?

25

A.

Yes.

A Google search estimates that probability by

898

Pazzani - direct
1

using

2

slide there.

3

Horling's deposition.

4

Q.

5

portion any different for the 27 -- these aspects of the

6

'276 patent versus the '040 patent?

7

A.

No.

8

Q.

And let's talk about Search Ads real quick.

9

put up 399 -- 292, please.

10

and the evidence is on the
PTX-0024, PTX-0385, PTX-0730, and Brian

And is your opinion on the estimated probability

Google infringes for the same reasons.
Can we

And we saw this before, just

what an ad is.

11

Is your opinion on the identification of the

12

properties the same for this element as it was for the

13

'040?

14

A.

15

of the ads.

16

Q.

17

please.

Yes.

So Google Search Ads identifies the properties
These properties are the

And let's skip a little bit.

.
Let's go to slide 295,

And this is PTX 401.

18

And what does this slide tell you about the

19

estimation of the probabilities aspect?

20

A.

Well, again, this is discussing

24

Q.

And let's turn to slide 296.

25

this slide?

21
22
23
And what is shown on

899

Pazzani - direct
1

A.

There is the evidence that I relied on for the

2

similar claim in '040.

3

PTX-0402, and PTX-0942.

4

Q.

5

please.

It's PTX-0115, PTX-0397, PTX-0400,

And let's go to Content Ads.

And can I get PTX-403,

Next slide.

6

And what is -- does Content Ad analyze

7

documents or identify properties of documents?

8

A.

Yes.

10

Q.

And let's go, let's skip to slide 301.

11

summarize the evidence you relied on for the, applying the

12

identified properties of the retrieved documents to the

13

user-specific learning machine to estimate a probability

14

aspect of this portion of the '276 patent?

15

A.

16

And the evidence for that is in PTX-0180, PTX-0222,

17

PTX-0400, PTX-0402, PTX-0408, PTX-0413, 1457, 1458, and

18

1462.

19

Q.

Content Ads are

9

Yes.

And can you

Google uses the short-term rephil clusters.

And let me have slide 302, please.

20

Can you summarize, is your opinion that each of

21

the accused Google products practice step, Step 1(f), the

22

retrieve document step of the '276 patent?

23

A.

Yes.

24

Q.

Is that what the green checkmark signifies?

25

A.

That's correct.

900

Pazzani - direct
1

Q.

And so let's talk about another element, slide 304.

2

And this is, in the '276, this is the using the estimated

3

probabilities for the respective plurality of retrieved

4

documents to present at least a portion of the retrieved

5

documents to the user.

6

Do you see that?

7

A.

Yes, I do.

8

Q.

And let me go to slide 305, please.

9

orient the jury what this step is about?
Yes.

Can you just

10

A.

This step is about after you've estimated the

11

probability that the user is interested in the document, how

12

you use that probability to show the user things they're

13

interested in.

14

Q.

15

evidence that you're relying on for step 1(f) of the '040

16

patent for this element of the '276 patent?

And are you relying on the evidence, the same

17

MR. VERHOEVEN:

18

THE WITNESS:

19

THE COURT:

20

Leading.

It's --

Hold on, Doctor.

I need to rule.

will sustain the question again.

21
22

Objection.

MR. NELSON:

I don't remember what the question

was.

23

THE COURT:

24

BY MR. NELSON:

25

Q.

Try your best.

So what -- can you identify what aspects of the

I

901

Pazzani - direct
1

Google Search, how the -- if the estimated probabilities are

2

used?

3

A.

4

earlier case because the estimated probabilities have to be

5

used to present portions of the documents to the user, so

6

the evidence will have, is a little bit different.

7

essence,

Yes, they are.

It's a little different than the

But, in

8
9

Q.

Right.

And I was talking about estimating the

10

probability aspect of it.

We'll talk about the presenting

11

to the user here in just a second.

12

A.

Okay.

13

Q.

So let's talk about that.

14

A.

Okay.

15

Q.

-- let's go to slide 308, please.

16

first.

So let's --

Let's do 307

17

So the rest of this element for 276 is -- one

18

back -- is to present at least a portion of the retrieved

19

documents to the user.

20

Can I have the next slide, please?

21

And the Court construed present as -- well, tell

22

the jury what the Court construed.

23

A.

To provide or make available.

24

Q.

And could I have slide 309, please.

25

of PTX-17.

And this is part

What does Google say about this element?

902

Pazzani - direct
1

A.

So this is part of the life of a query that describes

2

Google's web search system and it says that Google presents

3

them to the user, where then here it refers to the search

4

results.

5

Q.

6

of the search results make documents available for the

7

user?

8

A.

9

the documents on the web and they are made available to the

And how does the present -- does the presentation

Yes.

So the search results page contains links to

10

user because before those search results page existed, the

11

user might not have known how to access those documents, but

12

by analyzing the contents of the documents and then

13

displaying it on the search result page, they're now made

14

available to the user.

15

Q.

16

back to slide 306.

17

relied on for the, using the estimate probabilities aspect

18

of this claim?

19

A.

20

same as the '040 patent and this is Brian Horling's

21

testimony, PTX-0044, PTX-0039, and PTX-0200.

22

Q.

23

please.

24

probabilities step.

25

And let's talk about Search Ads next.

Yes.

And let me go

And can you tell me what evidence you

So using the estimated probabilities is the

And let me now jump to search address, slide 311,
And let's talk about the using the estimated

What evidence did you -- first of all, do you

Pazzani - direct

903

1

conclude that this portion of the step is met in the '276

2

patent?

3

A.

4

probabilities, that's done the same way as the '040 patent.

Yes.

5
6

With respect to using the estimated

For Search Ads, the evidence was PTX-0110, and
0403.

And the Gopalratnam deposition.

7

And for Content Ads it's also PTX-0110, but also

8

PTX-0223.

9

Q.

And just to the extent that not everything that you

10

stated is on these slides, are you also relying on the

11

testimony that you gave previously regarding these elements

12

for the '040 patent?

13

MR. VERHOEVEN:

14

THE COURT:

Objection, leading.

Overruled.

You can answer.

15

BY THE WITNESS:

16

A.

Could you ask the question again?

17

Q.

Sure.

18

every single exhibit or every single aspect of your

19

testimony regarding the '040 patent, are you also relying on

20

that portion of the testimony for the '040 patent that

21

relates to the respective '276 claim elements for part of

22

your opinion, the '276?

23

A.

24

my earlier testimony or my deposition or my report.

25

Yes.

To the extent that these slides don't contain

I think I rely on the same evidence that was in

MR. NELSON:

So let's turn to the presenting at

904

Pazzani - direct
1

least a portion of the retrieved documents to the user

2

aspect.

3

we zoom that one up a little bit?

4

BY MR. NELSON:

5

Q.

6

presenting advertisements to the user?

7

A.

Slide 312, please.

This is Part of PTX-17.

Yes.

8
9

That's not really legible.

Can

What is Google say about

It says -MR. VERHOEVEN:

Objection to the form of the

question.

10

THE COURT:

11

Counsel?

MR. NELSON:

I can rephrase it.

That's fine.

12

BY MR. NELSON:

13

Q.

14

presenting -- what does it say about ads to the user?

This is a Google document.

15
16

What does it say about

Let me just ask it again.

That was kind of

garbled.

17

Is PTX-17 a Google document?

18

A.

Yes, I believe so.

19

Q.

And can you read step 25?

20

A.

Yes.

21

Q.

What is the page that is being talked about there?

22

A.

The page is the search results page.

23

Q.

Is it also the advertising page?

24

A.

I'd want to see more of the document.

25

GFE presents the page to the user.

MR. NELSON:

Can you zoom back a little bit?

905

Pazzani - direct
1

BY THE WITNESS:

2

A.

3

querying the ad server.

4

are combined together on the page in step 25.

Yes, this is the part of the document that is

5

MR. NELSON:

So the ads and the search results

And let's talk about the Content

6

Ads aspect of this element.

7

probabilities.

8

BY MR. NELSON:

9

Q.

First, the estimating

Can I get slide 313, please?

And you have seen this slide before.

Is this the

10

same evidence, the estimating probability for Content Ads,

11

that you used for the '040 patent?

12

A.

13

should be ads are displayed in order of ad rank, highest at

14

the top and lowest at the bottom.

15

Q.

16

presentation of ads element of Content Ads?

17

opinion that that is met in '276 patent?

18

A.

Yes.

19

Q.

How is accomplished?

20

A.

Through some javascript in an I-frame that contains

21

the ads.

22

It's the same evidence but we highlighted it so it

What about the presentation?

What about the
Is it your

The ads are displayed on Content page.

MR. NELSON:

23

314.

24

BY MR. NELSON:

25

Q.

And let me turn to the next slide,

Can you summarize your opinion with respect to

906

Pazzani - direct
1

whether Element 1(g) is met by Google Search, Search Ads,

2

Content Ads, and YouTube?

3

A.

Yes, it is.

4

MR. NELSON:

It's a good time to stop.

I have

5

three dependent claims yet and I don't think I will finish

6

by 4:30, Your Honor.

7

MR. VERHOEVEN:

May I have a very brief sidebar?

8

THE COURT:

9

(Sidebar conference held.)

You may.

10

THE COURT:

Mr. Verhoeven.

11

MR. VERHOEVEN:

I don't want to keep the jury

12

late.

I was thinking I might ask if he would finish it, but

13

I don't want to keep the jury late.

14

My request would be that they not talk to the

15

witness about the stuff they have already covered just as if

16

they finished direct.

17

about the slides they haven't covered, I have no objection

18

to that.

19

is okay.

If they want to talk to the witness

And then we'll just pick it up tomorrow, if that

20

THE COURT:

21

MR. NELSON:

What is your position?
I may need to talk to the witness a

22

little bit because I need to go back through, I've got a

23

list of exhibits that I believe have been entered into

24

evidence, and I've got holes in it.

25

I've been trying to keep track as closely as I

907
1

could as to what got in and what didn't but to the extent

2

there is evidence for Content Ads or Search Ads that there

3

is nothing written there, at least one, if I could figure

4

that out and do that.

5

MR. VERHOEVEN:

6

MR. NELSON:

7

MR. VERHOEVEN:

There is --

Direct is still open.
There is a written transcript,

8

and everything that has been admitted has been admitted.

9

would submit counsel could do that without counselling the

10

I

witness.

11

THE COURT:

12

MR. NELSON:

Well, are you -I'm not done.

I'm not asking to be

13

able to coach the witness or something, but direct is still

14

open and I need to go back through and make sure that I have

15

everything I need.

16

might need to ask him about that might have been missed.

17

was trying to check everything off but I don't think it's

18

fair to me to not about able to ask him a question about

19

something that might have been missed on direct that I want

20

to follow-up on.

I think I do, but if something is up I

21

THE COURT:

22

MR. VERHOEVEN:

I

Is that your request?
My concern, Your Honor, is we're

23

going to get another hour and-a-half when we would have had

24

ten minutes because they're going to go back and work on it

25

again tonight.

And whatever Your Honor wants, but I'm

908
1

trying to move it along.

2

wanted to talk about what they have left, that I wouldn't

3

object to that.

4

And I was thinking that if they

If all you are going to do is, you represent,

5

just talk about what exhibits you have in or out, you won't

6

talk about any substantive issues, you represent that, then

7

I think --

8

MR. NELSON:

9

THE COURT:

I can't represent that 100 percent.
That's fine.

I understand the

10

concern that Mr. Verhoeven is raising but the reality is

11

direct is still open.

12

time today.

13

Plaintiff has used an awful lot of

The jury has been here, been through a lot, so

14

I'm not going to keep the jury for however many minutes are

15

left.

16

remaining minutes anyway.

17

and-a-half on cross tomorrow, I mean more direct, he is

18

going to use an hour and-a-half more, and you are going to

19

have a full opportunity to cross.

20

used up the whole day on direct, his witness is still on

21

direct.

22

talked about last night but they were free to do so.

23

They're free to do so tonight.

24

tomorrow.

25

direct so I'm going to allow him to confer with his client.

He is not going to finish his direct in those
If he comes back with an hour

It's just like yesterday.

But the fact is he has

I don't know what they

And we'll see what happens

I understand the concern but the witness is on

909
1

MR. VERHOEVEN:

2

THE COURT:

3

Understood, Your Honor.

All right.

So we will stop now.

I'll tell the jury to go.

4

MR. NELSON:

Thank you, Your Honor.

5

THE COURT:

6

(Sidebar conference ends.)

7

THE COURT:

Okay.

Ladies and gentlemen, we will let

8

you go at what is maybe three minutes earlier than our

9

deadline.

10

Of course, no talking about the case while you

11

are gone.

Don't do any research or read anything about the

12

case.

13

be here in time to do that.

We expect to start at 9:00 o'clock tomorrow so please
Have a very good evening.

14

(Jury left courtroom.)

15

THE COURT:

16
17
18
19
20
21

Doctor, you may to step down.

The

rest of you can have a seat.
I need to give you ruling on designations which
may be played tomorrow.
So first we have objections to the deposition
designations of the witness Roy Twersky.
There are some objections from both sides.

22

Some from PUM and some from Google.

23

objections are overruled with the following two exceptions:

24
25

All of both sides'

First, the Court sustained PUM's objection to
Google's designation at page 449 of Mr. Twersky's

910
1

deposition.

2

a risk of that excerpt being misleading to the jury as it

3

suggests I think incorrectly that the witness changed his

4

testimony within just the most recent five minutes before

5

being asked the question when that is not what appears to

6

have happened.

7

And in the Court's view, there is too great of

And we also are sustaining Google's objection

8

to PUM's counterdesignation at page 451 given that PUM's

9

counterdesignation is untimely, and there is no reason to

10

allow for that untimely counterdesignation under the

11

circumstances.

12

Beyond those two sustained objections, the

13

remainder of the objections with respect to Mr. Twersky are

14

overruled.

15

proper counterdesignations for completeness.

16

The others we felt at a general level were

The issue of Mr. Twersky's change of testimony

17

is clearly going to be important to Google's breach of

18

contract claim.

19

changed his testimony, including his relative role and level

20

of knowledge about the date of conception, compared to his

21

co-inventor Dr. Konig are all relevant to his credibility

22

and won't be unfairly confusing to the jury.

23

The different interpretations as to why he

Also, I think it is important to keep in mind

24

that to some extent at least this is a truth-seeking process

25

that we are engaged in and I think it's more likely that the

911
1

jury will come to a finding based on the truth if they do

2

hear all of what has been designated, counterdesignated with

3

respect to witness Mr. Twersky other than the two objections

4

that I have sustained.

5

A few specifics I wanted to comment on.

The

6

objections to the designations at pages 122 to 124 and 130

7

to 131.

8

his patent and what he thinks of the scope of his invention

9

did not seem to us to be highly probative of anything

The dispute as to how recently Mr. Twersky had read

10

relevant in this case, but similar questions were asked

11

and answered of co-inventor Dr. Konig live here at trial

12

sometimes even without objection and there are no Rule 403

13

factors that would favor excluding that testimony.

14

And with respect to the testimony designated at

15

page 126, Mr. Twersky's recollection as to whether or not he

16

and Dr. Konig had a working system is in the Court's view

17

relevant to the issue of the problem the patentee's thought

18

that they were solving.

19
20
21

So that is the rulings with respect to witness
Twersky.
With respect to witness Frank Montes, we have

22

Google's objections to PUM's counterdesignations.

23

objections are sustained.

24

22 is improper and not necessary for completeness.

25

Google's

PUM's counterdesignation at page

The witness's reference to the existence of an

912
1

SRI/Google agreement does not make it necessary to

2

counterdesignate testimony about discussions and invoices.

3

Similarly, PUM's counterdesignations of pages 29

4

to 30 and 42 is improper and not necessary for completeness.

5

Google's designated portion of the witness is simply

6

testifying there is an agreement that was acceptable to SRI.

7

PUM seeks to use that to counterdesignate about what SRI

8

knew or what SRI did or what Google inquired about and none

9

of that is necessary for completeness and is an improper

10

counterdesignation.

11
12

So that is the rulings on I think the issues in
front of me.

13
14

Is there anything we should talk about from
plaintiff's perspective?

15

MR. NELSON:

16

THE COURT:

17

MR. VERHOEVEN:

18

No, Your Honor.
How about for defendant?
I just need to confer for one

second.

19

THE COURT:

I can hope.

20

(Counsel confer.

21

MR. PERLSON:

While you wait, we can deal with

22

the Konig exhibits in the morning?

23

THE COURT:

24
25

Right.

tomorrow morning.
MR. PERLSON:

I can hope.

Okay.

Is that fine?

I'll be available at 8:30

913
1
2

THE COURT:

And Dr. Konig is not going to be on

the stand right away.

3

MR. PERLSON:

It seems like it.

4

THE COURT:

5

Is there an issue?

6

MR. VERHOEVEN:

I would be surprised.

No, there is not.

We've worked

7

it out.

8

these directs, and we're not going to object that they're

9

waived or anything.

10

We had not received the animations that went with

We're just trying to get them and I was

just told they were being sent to our hotel right now.

11

THE COURT:

I will consider it a nonissue then.

12

All right.

Well, we will look for you at 8:30

13
14

tomorrow morning.

Have a good night.

(Proceedings adjourn at 4:34 p.m.)

15
16
17
18
19
20
21
22
23
24
25

I hereby certify the foregoing is a true and accurate
transcript from my stenographic notes in the proceeding.
/s/ Brian P. Gaffigan
Official Court Reporter
U.S. District Court