Personalized User Model LLP v. Google Inc.

Filing 502

Letter to The Honorable Leonard P. Stark from Regina S.E. Murphy regarding Amended Exhibit L to Declaration of Jennifer D. Bennett (D.I. 454) - re 454 Declaration,. (Attachments: # 1 Exhibit L)(Murphy, Regina)

Exhibit L

Page 1
1
2
3
4

IN THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF DELAWARE
PERSONALIZED USER MODEL, LLP,
Plaintiff,

5
v.
6
7
8

GOOGLE INC.,
Defendant.
______________________________
GOOGLE INC.,
Counterclaimant,

9
v.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

PERSONALIZED USER MODEL, LLP
and YOCHAI KONIG,
Counter Defendants.

)
)
)
) 09-525-LPS
)
)
)
)
)
)
)
)
)
)
)
)

DEPOSITION OF MUSTAFA KEMAL SONMEZ
Taken on behalf of Defendant and Counterclaimant
* * *
BE IT REMEMBERED THAT the deposition of
MUSTAFA KEMAL SONMEZ was taken before Rosemary
Tanzer, a Registered Professional Reporter and a
Certified Shorthand Reporter for Oregon and
Washington, on Monday, August 15, 2011, commencing
at the hour of 9:24 a.m., at LNS COURT REPORTING
AND CAPTIONING, 1123 S.W. Yamhill Street, Portland,
Oregon.
Job No. CS344933
Veritext Corporate Services
800-567-8658

973-410-4040

Page 34
1
2

MR. FRIEDMAN:
no.

No, no, no, no, no,

Over the next few weeks.

3

THE WITNESS:
BY MS. ROBERTS:

Okay.

4

Q

So then the -- the speaker

5

verification project, did it use models or modeling

6

techniques?

7

A

8

techniques.

9

particular project.

Everything uses models and modeling
It's not a distinction about that
But yeah, like anything that

10

uses computers, I suppose you need to build a model

11

for, so yeah.

12

Q

Now --

13

A

Just a general statement, I don't know what --

14

it wouldn't really say much about the particular

15

problem or task.

16

Q

17

verification, would there have been -- I guess,

18

were the models connected to particular speakers or

19

to particular sounds?

20

A

21

but, yes.

22

develop a model per speaker.

23

are other approaches where you would sort of

24

transform the features into a space and

25

discriminate over there, so that wouldn't

Now, for the particular task of speaker

Well, there are different ways of doing that,
Some actually -- some approaches would
And then -- but there

Veritext Corporate Services
800-567-8658

973-410-4040

Page 35
1

necessarily be separate models for speakers.

But

2

yeah, there are algorithms that do both.

3

Q

4

at SRI, was there any particular approach you were

5

using or were you using --

6

A

7

actually, essentially, was an amalgam of, like,

8

maybe six or seven different systems.

9

example, Yochai's front end would be one of those

10

systems we would combine to get better accuracy.

11

And yes, some of those systems did have a model per

12

speaker.

13

Q

14

networks"?

15

A

Yes.

16

Q

Were neuro networks used in this speaker

17

verification project?

18

A

19

technique involves estimating a neuro network

20

and -- but, also, that's, again, a very general

21

thing because neural networks also kind of have

22

come to refer to things as general as machine

23

learning or statistics.

24

I would say.

25

Q

In terms of the speaker verification project

Well, we use many approaches.

And our system

Like, for

Are you familiar with the term "neuro

They were.

Right.

Actually, Yochai's

That's a very broad area,

Are you familiar with Bayesian learning
Veritext Corporate Services

800-567-8658

973-410-4040

Page 55
1

2000, there was a meeting recognition project that

2

Mr. Stolcke worked on?

3

A

Yes.

4

Q

And when you say after 2000, do you recall

5

roughly when that would have been?

6

A

7

2004, although I'm not positive.

8

Q

9

work in machine learning outside the context of

I can guess.

I want to say maybe 2003 or

In the 1996 to 1999 time frame, did SRI do any

10

speech recognition or speaker verification?

11

A

12

bioinformatics could be an exception, but like

13

99 percent was really speech and speaker.

14

that the STAR Lab did related to that.

15

Q

16

Lab but SRI more broadly, do you know whether it

17

did any work in machine learning?

18

A

19

framework.

20

computers at SRI would be doing some sort of

21

machine learning.

22

Q

Did SRI do any work in personalization?

23

A

I don't know what that means, personalization.

24

Q

Did it do any work in providing personalized

25

computer services to computer users?

I guess -- let me see.

So let me clarify.

Oh, yeah.

My work on

Anything

Not limited to the STAR

Machine learning is a very general

So anybody who is working on -- with

Veritext Corporate Services
800-567-8658

973-410-4040

Page 56
1

A

That's certainly not within STAR Lab.

There

2

was nothing like that.

3

don't know.

4

Q

Have you ever heard of the CALO project?

5

A

Yes, I have.

6

Q

What is that?

7

A

CALO was essentially -- let's see.

8

person who led that project, Adam Cheyer,

9

C-H-E-Y-E-R.

In terms of general SRI, I

I know the

So CALO is the project that -- so

10

what I know is it's a very large project, first of

11

all.

12

of -- they essentially -- CALO must be DARPA, I

13

think.

14

like almost every single researcher in the field.

15

And they picked SRI as, like, an integrator of

16

many, many different technologies.

17

of those, but there was all kinds of other

18

technologies, essentially, like image, documents,

19

everything, I think, all under one roof.

SRI was like an integrator of many, many sort

20

DARPA funded the whole field, pretty much,

Speech was one

So SRI's task was to integrate all of

21

those technologies into a single system in an age

22

of framework which Adam Cheyer had developed, which

23

is, essentially, you have these independent agents

24

that each working on -- one working on speech, one

25

goes to, who knows, maybe e-mails, and one goes
Veritext Corporate Services
800-567-8658

973-410-4040

Page 58
1

Q

BY MS. ROBERTS:

In your time at SRI, did it do

2

any work in Internet search?

3

MR. FRIEDMAN:

Objection as to form.

4

A

I don't know anybody who did Internet search

5

at SRI.

6

Q

BY MS. ROBERTS:

7

A

I mean, I can't be sure about the whole SRI,

8

but in the STAR lab certainly that was not a topic

9

of research.

At SRI.

And among the projects I know that's

10

like the AI center, which was a close related

11

artificial intelligent center, I don't think

12

anybody was doing research over there either.

13

I can only be sure about the STAR Lab.

14

Q

15

wouldn't know for sure one way or another?

16

A

17

project that I know of, but I can't say there was

18

none, other than the speech for the STAR Lab.

19

Q

20

intelligence lab; is that correct?

21

A

Right.

22

Q

Is that a lab that you worked with in

23

connection with your work at the STAR Lab?

24

MR. FRIEDMAN:

25

Q

But

In terms of other labs or divisions, you just

I can't be sure.

Okay.

There is not a research

Now, you mentioned the artificial

BY MS. ROBERTS:

In what time frame?

We'll start with the entire time

Veritext Corporate Services
800-567-8658

973-410-4040

Page 59
1

frame.

2

A

3

is, again, relates to my work in computational

4

biology.

5

called Bio-SPICE for which I was the proposal

6

manager.

7

Peter Karp, who was in the AI center.

8

computational biologist.

9

common projects that I have with the AI lab.

Right.

So my involvement with the AI center

There was a very large DARPA program

And a very central figure in that was
He's a

I can't recall any other
AI

10

lab is just like another lab, just like speech

11

technology and research lab, a separate entity

12

within the same information technology division.

13

Q

14

you didn't interact with the AI lab?

15

A

No.

16

Q

Is it AI lab or AI center?

17

A

I don't know.

18

a lab too.

That's my guess.

19

Q

If I call them the lab, you'll

20

understand who I'm referring to?

21

A

Yes.

22

Q

Okay.

23

believe that SRI would not do any work in Internet

24

search?

So then, for the speaker verification project,

Okay.

25

Doesn't matter?

We were a lab, so they must be

As of 1999, did you have any reason to

MR. FRIEDMAN:

Objection as to form.

Veritext Corporate Services
800-567-8658

973-410-4040

Page 60
1

A

I know for a fact that speech technology and

2

research lab was not doing any re -- because I know

3

all the projects involved, and Internet search is

4

really something very different.

5

In terms of timing, I'm not even sure.

6

Is that before Google?

7

Internet search was even hot then.

8

long time ago.

9

Q

BY MS. ROBERTS:

10

A

But I know that STAR Lab was not doing

11

anything related to Internet search.

12

Q

13

the SRI as a whole, you don't know one way or

14

another?

15

A

16

I knew of that did Internet search, but I can't say

17

nobody else was.

18

Q

19

Internet search at that time; is that correct?

Okay.

I don't even know if
It's such a

Right.

And in terms -- beyond the STAR Lab,

I can't be sure.

There was not a project that

And you're not sure if you were even aware of

20

MR. FRIEDMAN:

Object to the form.

21

Q

BY MS. ROBERTS:

Do I understand you correctly?

22

A

I'm just wondering because it's such a long

23

time ago.

24

remember when, actually, I started using Google,

25

for example.

For example, I honestly cannot -- cannot

I just don't recall.

But it can't be

Veritext Corporate Services
800-567-8658

973-410-4040

Page 61
1

that far away from that time.

So I think Internet

2

search really -- I'm just guessing, it must have

3

taken -- it must have taken off much after that,

4

I'm guessing.

5

becoming an issue or something that got discussed

6

at that time.

7

speech and language.

8

Q

9

1999, did you have any reason to believe that SRI

But I don't remember ever even that

It was -- we focused solely on

And did you -- in the same time period, as of

10

would not do any work in providing personalized

11

computer services to users?

12

MR. FRIEDMAN:

Objection as to form.

13

A

Well, I know again, speech technology and

14

research lab was not providing any sort of

15

personalized user experience or -- because that

16

wasn't really what we were -- we didn't really have

17

users, so to speak.

18

systems that was meant for processing of large

19

corpora of data.

20

corpus versus corpora.

21

speech algorithms.

22

for sure for STAR Lab, there wasn't any of that.

23

SRI, since it's bigger, even though there is not a

24

project that I was aware of, I can't say there

25

wasn't any.

It was really more large

That's C-O-R-P-O-R-A, like a
So it's really to develop

So we weren't -- so I can say

Veritext Corporate Services
800-567-8658

973-410-4040

Page 62
1

Q

BY MS. ROBERTS:

Are you familiar with the

2

company called Utopy?

3

A

Yes.

4

Q

What is that company?

5

A

So Utopy is the company that Yochai Konig

6

founded after he left SRI.

7

Q

And what kind of work does Utopy do?

8

A

Well, again, I didn't follow this very

9

closely, but it started out as a search company, a

10

personalized search company.

I know this because

11

after Yochai left, like maybe a year or maybe two

12

years after he left, I was -- I was invited to be a

13

Beta user on Utopy's first version.

14

and I took a quick look, but I didn't have much

15

time, essentially.

16

passed, I heard that they were now doing more

17

speech analytics.

I logged in

But then, after some time

18

(Reporter requests clarification.)

19

THE WITNESS:
BY MS. ROBERTS:

Analytics.

20

Q

21

learned about Utopy?

22

A

23

year or -- it was really when I received this Beta

24

user invitation.

25

Q

Yes.

Do you recall when you first

So that was -- that must be, like, a

And you said that was roughly a year or two
Veritext Corporate Services

800-567-8658

973-410-4040

Page 63
1

after Mr. Konig left SRI?

2

A

3

the time.

4

Q

5

later than that?

6

A

7

can't be sure.

8

least a year after he left, I think, because we

9

didn't hear from him.

I'm just guessing.

I don't know exactly when

So maybe in the 2000/2001 time frame, but not

That must be -- that sounds right, although I
The thing is, it was -- it was at

And then -- so I just

10

remember the name Utopy, which sounded like an

11

interesting name.

12

Q

13

this personalized search product?

14

A

That's right.

15

Q

And you didn't ultimately --

16

A

Well, I mean, I logged in.

17

idea was -- it was essentially making a model of

18

the type of searches that you do such that -- and

19

then use this profile to do further searches.

20

That's what I got the idea was.

21

said, at that point, I think -- I don't know, maybe

22

I'm a little -- the Internet search was not like a

23

huge part of life as it is now in some ways.

24

wasn't doing Internet searches every single day, I

25

don't think.

I don't know.

And so you were invited to be a Beta user of

And I guess the

And -- but like I

So I

Veritext Corporate Services
800-567-8658

973-410-4040

Page 66
1

thing came, essentially, as invitations, meant

2

nothing to the lab, because it was something

3

different, completely, from what the lab was

4

working on.

5

found something else to do, that kind of thing.

6

Q

7

you?

8

A

It doesn't.

9

Q

The program that you were invited to be a Beta

So it was actually -- so they had just

Does the name Personal Web ring a bell for

10

user of, do you recall if it had a name?

11

A

12

website was Utopy, but I don't recall the name of

13

the program.

14

Q

15

program?

16

A

17

meeting or -- so I don't know if somebody at SRI --

18

when or how they became -- if they became aware.

19

But I myself did not become aware until I received

20

the invitation.

21

whole, when they became aware.

22

Q

23

user of this program, did you have any other

24

contact with Mr. Konig after he left SRI?

25

A

It was called Utopy, I think.

Like the

Do you know whether SRI was ever aware of that

It was never, like, either mentioned in a

So I don't know if SRI, as a

Now, other than being invited to become a Beta

Well, we may have hung out.

We may have spent

Veritext Corporate Services
800-567-8658

973-410-4040

Page 90
1

rights to and that Yochai and others are -- I don't

2

know this third person at all, Michael Berthold.

3

And then really the question involved what Yochai

4

was working on before he left SRI and whether there

5

is any reasonable way in which somehow these

6

patents could be related to the work he was doing

7

while he was at SRI.

8

said, essentially, Internet search or anything

9

related to that was not really the lab's -- the

10

And again -- and again, I

lab's focus or topic in any way.

11

And also we discussed, for example, just

12

the fact that machine learning or computer science

13

could be used in both.

14

know, I said that doesn't -- that those are such

15

general frameworks and tools that it really does

16

not, in my view, provide a link between what he was

17

doing before he left SRI and what he did with

18

Utopy, essentially.

19

me and I gave an answer.

20

Q

21

patents provided to you?

22

A

23

the first time I'm seeing them.

24

specific claims or anything like that.

25

Q

In my view essentially, you

That was -- that was asked to

Was an explanation of what's covered in these

No.

We didn't look at the patents.

These are

We did not discuss

Did you discuss any specific issues in the
Veritext Corporate Services

800-567-8658

973-410-4040