Personalized User Model LLP v. Google Inc.

Filing 608

REDACTED VERSION of 585 Reply Brief, re 555 MOTION TO EXCLUDE PORTIONS OF DR. EDWARD FOX'S NON-INFRINGEMENT REPORT FOR FAILURE TO APPLY THE COURT'S CLAIM CONSTRUCTION by Personalized User Model LLP. (Tigan, Jeremy)

IN THE UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF DELAWARE
PERSONALIZED USER MODEL, L.L.P.,

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

C.A. No. 09-525 (LPS)
REDACTED - PUBLIC VERSION

PUM’S REPLY BRIEF IN SUPPORT OF ITS MOTION TO EXCLUDE
PORTIONS OF DR. EDWARD FOX’S NON-INFRINGEMENT REPORT FOR
FAILURE TO APPLY THE COURT’S CLAIM CONSTRUCTION
OF COUNSEL:
Marc S. Friedman
SNR DENTON US LLP
1221 Avenue of the Americas
New York, NY 10020-1089
[phone]
Mark C. Nelson
SNR Denton US LLP
2000 McKinney Avenue, Ste. 1900
Dallas, TX 75201
[phone]
Jennifer D. Bennett
SNR DENTON US LLP
1530 Page Mill Road, Ste. 200
Palo Alto, CA 94304-1125
[phone]

Original Filing Date: February 19, 2014
Redacted Filing Date: February 28, 2014

MORRIS, NICHOLS, ARSHT & TUNNELL LLP
Karen Jacobs Louden (#2881)
Jeremy A. Tigan (#5239)
Regina Murphy (#5658)
1201 N. Market Street
P.O. Box 1347
Wilmington, DE 19899-1347
[phone]
[e-mail]
[e-mail]
[e-mail]
Attorneys for Personalized User Model, L.L.P. and
Yochai Konig

TABLE OF CONTENTS
Page
TABLE OF AUTHORITIES ...................................................................................................... ii
I.

INTRODUCTION ........................................................................................................... 1

II.

THE COURT ALREADY REJECTED DR. FOX’S CONSTRUCTION
THAT THE USER MODEL MUST BE “SEPARATE,” “EXCLUSIVE,”
OR “HIS OWN” .............................................................................................................. 2
A.
B.

III.

The Court Addressed the Dispute Regarding the “Specificness”
Issue..................................................................................................................... 2
Google Cites Irrelevant Evidence to Confuse the Issue. ........................................ 4

THE COURT SHOULD EXCLUDE DR. FOX’S OPINIONS THAT
“ESTIMATING” CANNOT BE SATISFIED BY A CALCULATION OR
COMPUTATION ............................................................................................................ 5
A.

Google Ignores the Statements in Dr. Fox’s Report .............................................. 5

B.

Google Again Raises Irrelevant Issues to Avoid the Real Issue. ........................... 9

IV.

THE COURT SHOULD EXCLUDE DR. FOX FROM TESTIFYING
THAT PROBABILITY MUST BE BETWEEN O AND 1. .............................................. 9

V.

THERE IS NO BASIS TO EXCLUDE ANY OF DR. CARBONELL’S
OPINIONS. ................................................................................................................... 10

VI.

CONCLUSION ............................................................................................................. 10

-i-

TABLE OF AUTHORITIES
Page(s)
CASES
Cook Inc. v. Endologix, Inc.,
No. 1:09-cv-01248-TWP-DKL, 2012 WL 3886204 (S.D. Ind. Sept. 6, 2012) ........................ 4
Insight Technology, Inc. v. SureFire, LLC,
2009 WL 3242557 (D.N.H. 2009) ......................................................................................... 4
MarcTec, LLC v. Johnson & Johnson,
No. 07-CV-825, 2010 WL 680490 (S.D. Ill. Feb. 23, 2010)................................................... 4
Saffran v. Johnson & Johnson,
Civil Action No. 2:07–CV–451, 2011 WL 197871 (E.D. Tex. Jan. 20, 2011) ........................ 4
OTHER AUTHORITIES
Local Rule 7.1.2 ........................................................................................................................ 10

- ii -

I.

INTRODUCTION
Under the guise of noninfringement, Google and Dr. Fox try to reconstrue the claims.

Google concedes as much, arguing that PUM should not be able “to prevent Google’s
noninfringement expert from interpreting the claims in the same manner as PUM’s validity
expert.” Opp., D.I. 576, at 2 (emphasis added). Claim interpretation, however, is reserved for
the Court. The Court rejected Google’s attempt to require that the “User Model specific to the
user” and “user-specific learning machine” be “unique,” “restricted,” or the newly-minted
“exclusive”1 to the user. (D.I. 347, at 27) (noting that “Google’s use of the word ‘unique’ is
inconsistent with the patent’s specification”). Nonetheless, throughout his report Dr. Fox opines
that each user must have his “own” model (e.g., the butler and car hypotheticals in paragraphs
318 and 319 of the Fox Report). The Court should reject this already-decided argument and
exclude the portions of the Fox Report that rely on it.
Dr. Fox’s opinions regarding the “estimating” limitations also fail to follow the Court’s
claim construction. During claim construction, Google argued that “estimating” is limited to
calculating. The Court disagreed and found the term to be broader. The Court did not state that
estimating could not be satisfied by a calculation; only that the term is not limited to calculation,
as Google had argued. (D.I. 347, at 32-33). Dr. Fox’s report ignores that finding, stating “[t]he
Court explicitly rejected that ‘estimating’ could be met by a calculation.” E.g., Fox Report
¶¶ 290-292, 498, 521 (Ex. 1, to D.I. 557). Google now tries to salvage Dr. Fox’s opinions by
arguing that “estimating” precludes “precise” calculations. Opp. at 8-10. But that is not what
Dr. Fox’s report states; Dr. Fox does not analyze the “precision” of the accused calculations.

1

See, e.g., Fox Report at ¶¶ 318 and 319, attached as Ex. 1 of the Bennett Declaration
(D.I. 557).

Regardless, however, Google’s “precision” argument is not based on the Court’s
construction which did not impose any sort of “precision” requirement on the claims—
“‘estimating’ means ‘approximating or roughly calculating.’” (D.I. 348, at 2). The claim
language that is being referred to is “estimating” the “parameters” of a learning machine and of
the “probability” that the user is interested in a document. These values, regardless of precision,
are estimates (i.e., approximations or rough calculations) because they utilize observed
information in an attempt to predict values that by their very nature are not knowable with 100%
accuracy (e.g., the user’s interest in an unseen document). Dr. Fox’s attempt to manufacture a
non-infringement argument by misapplying the Court’s construction should be rejected. The
Court should exclude Dr. Fox’s opinions that rely on this incorrect construction.
Lastly, Google attempts to confuse the issues by citing to PUM’s expert reports,
depositions, and the reexaminations. But the only issue raised by PUM’s Motion was whether
Dr. Fox correctly applied the Court’s claim construction.

PUM’s expert reports, expert

depositions, and reexamination filings are not germane to whether Dr. Fox’s opinions are based
on the Court’s construction.

Moreover, as set out below, these PUM materials are not

inconsistent with the Court’s claim construction.
II.

THE COURT ALREADY REJECTED DR. FOX’S CONSTRUCTION THAT THE
USER MODEL MUST BE “SEPARATE,” “EXCLUSIVE,” OR “HIS OWN”
A.

The Court Addressed the Dispute Regarding the “Specificness” Issue.

Google incredibly argues that “the Court did not squarely address whether the User
Model [or user-specific learning machine] needed to be user specific in the corresponding
construction.” Opp. at 7 n.4. To the contrary, the Court acknowledged there was a dispute
regarding “specific” and resolved it in PUM’s favor.

-2-

The Court rejected the notion that

“specific” means that each user must have his “own” user model or learning machine. For
starters, the Order acknowledged that the parties disputed the meaning of “specific”:
The crux of the dispute is what the patent means when it says “specific” or
“specific to the user.” (Tr. at 61) To resolve this dispute, the Court must
determine (1) whether the User Models must be “unique,” as Google suggests and
(2) whether the User Model must be stored in a data structure.

D.I. 347 at 24-25.

The Court also stated, “Resolution of this dispute turns on whether a User

Model is specific because it has completely different variables than other User Models, or if,
instead, a User Model is specific because it has completely different numerical values than other
User Models.” Id. at 26 (underlining added).
Google fails to explain how the Court nonetheless failed to address the “specificness”
issue. Google’s counsel, in fact, argued the exact issue that Dr. Fox now attempts to re-raise:
“And this really is the crux of the dispute. What our construction provides is that each
individual user has their own user model. I have a user model. You have a user model. Other
people, they each have their own user model.”

(Hr’g Tr., D.I. 170, at 61, Jan. 11, 2011

(emphasis added).) Page 61 is the same page the Court cited in rejecting Google’s construction.
(D.I. 347, at 24). Google also ignores the Court’s discussion of the specification addressing this
issue: “Prototype users, and the fact that a User Model can be initialized without any userspecific information, undercuts Google’s argument … Google’s attempt to harmonize this
embodiment of the patent with its proposed construction was unpersuasive.” (D.I. 347, at 26).
Dr. Fox nonetheless advances the very argument the Court rejected, asserting that “the
claim requires that each user have his own User Model, i.e., a User Model specific to the user.”
(D.I. 557, at ¶ 318). Google does not even try to distinguish Dr. Fox’s argument from Google’s
rejected construction, and instead ignores the transcript citation entirely. And Google does not—
because it cannot—argue that there is any difference between the “separate,” “exclusive,” or

-3-

“own” requirements set forth in Dr. Fox’s report and Google’s previously rejected “unique” and
“restricted to” claim construction arguments.
Google’s citations to Dr. Fox’s report, moreover, demonstrate that Dr. Fox is performing
claim construction, rather than applying the Court’s construction. Google cites the part of the
report that states, “The plain language of the claims requires that each user have his own User
Model.” Opp. at 5 (first emphasis added). Not only is that wrong, but more importantly, the
Court’s construction of the disputed term—not the “plain language”—controls. Google also
argues that Dr. Fox should be allowed to “interpret[] the claims in the same manner . . .” Opp. at
2. But claim construction is for the Court.
Lastly, Google argues that “PUM’s motion miscasts the Court’s Markman Order as
somehow ruling that there need be no user-specificity at all.” Opp. at 4. That is incorrect. The
Court’s construction, which PUM’s experts apply, requires the User Model be “specific” to the
user (and that there be a “user-specific” learning machine (‘276 patent)). As was explained
during claim construction, it is the parameters estimated from the user-specific data files that
make the User Model/learning machine “specific” to the user. (D.I. 170, at 117:7-119:23).
Dr. Fox’s contrary opinions should be stricken.2
B.

Google Cites Irrelevant Evidence to Confuse the Issue.

Trying to distract the Court’s attention from the issue—whether Dr. Fox followed the
Court’s claim construction—Google cites to PUM’s expert reports, expert depositions, and a
reexamination response. Opp. at 7-8. Although irrelevant to the propriety of Dr. Fox’s opinions,
2

See, e.g., Cook Inc. v. Endologix, Inc., No. 1:09-cv-01248-TWP-DKL, 2012 WL 3886204, at
*3 (S.D. Ind. Sept. 6, 2012); see also MarcTec, LLC v. Johnson & Johnson, No. 07-CV-825,
2010 WL 680490, at *4 (S.D. Ill. Feb. 23, 2010); Saffran v. Johnson & Johnson, Civil Action
No. 2:07–CV–451, 2011 WL 197871, at *3–4 (E.D. Tex. Jan. 20, 2011); Insight Technology,
Inc. v. SureFire, LLC, 2009 WL 3242557, *1–*2 (D.N.H. 2009).

-4-

none of these materials are inconsistent with the Court’s construction. For example, Google
quotes (out of context) paragraphs 246 and 373 of the Carbonell report. Opp. at 7. But those
paragraph are not directed to “User Model specific to the user.” Instead, Dr. Carbonell addressed
the claimed requirement that parameters are estimated from user-specific data files: “Joachims
does not teach estimating parameters of a learning machine wherein the parameters are estimated
in part from the user-specific data files” (Google Ex. 2 ¶ 373 (emphasis in original)); and “this
limitation requires that the parameters be based in part on the user-specific data files” (Google
Ex. 2 ¶ 246 (emphasis in original)). Similarly, the reexamination responses both addressed
“parameters of a learning machine estimated from user-specific data files.” (Sistos Decl. Ex. 7 at
19 (emphasis in original); see also Sistos Decl. Ex. 8 at 22-23.)3
The expert depositions do not help Google either. Opp. at 7-8. Both Dr. Pazzani and
Dr. Carbonell stated that some of the data must be specific to the user. Sistos Ex. 5 at 264:16265:3; Sistos Ex. 4 at 39:18-40:14. But that is not what Dr. Fox is saying. He is saying that the
variables—not the data—must be specific to the user, and the Court rejected that argument.
III.

THE COURT SHOULD EXCLUDE DR. FOX’S OPINIONS THAT
“ESTIMATING” CANNOT BE SATISFIED BY A CALCULATION OR
COMPUTATION
A.

Google Ignores the Statements in Dr. Fox’s Report

Google’s opposition ignores the Fox Report. PUM’s Motion identified 22 paragraphs
from the Fox Report that argue that estimating cannot be satisfied by “calculating” or
“computing.” Mot., D.I. 556, at 8-9. Google’s Opposition only cites to two of those paragraphs
and does not address the specific language in any of them.

Opp. at 2, 8-10. Instead of

addressing the Report, Google argues that “PUM relies on a number of out-of-context quotes.”
3

In addition, the 2011 reexamination responses — which are irrelevant to this
litigation — pre-date the Court’s 2012 claim construction.

-5-

Opp. at 9. There is nothing out of context—a heading in the Report states that “E. Calculating is
not estimating.” (D.I. 557, Ex. 1, Page 133 (emphasis in original)). The remainder of Fox’s
contains similar statements as set forth in the table below.
ASSERTION IN GOOGLE’S
BRIEF

QUOTES FROM DR. FOX’S REPORT4

PUM also argues Dr. Fox’s
opinions as to the terms “estimating
parameters of a learning machine”
and “estimating a probability that a
document is of interest to user”
should be excluded because he
supposedly opines that estimating
“excludes” calculating.
But
Dr. Fox advances no such opinion.
(Opp. at 2 (emphasis added).)

The Court explicitly rejected that “estimating” could be
met by a calculation (¶ 290)
The Court explicitly rejected that “estimating” could be
met by a computation (see above). (¶ 292)

and the Court explicitly rejected construing “estimating”
as “calculating.” (¶ 405)

and the Court explicitly rejected construing “estimating”
as “calculating.” (¶ 464)

. . . . . The Court explicitly rejected that
“estimating” could be met by a computation . . .” (¶ 498)

and the Court explicitly rejected construing
“estimating” as “calculating.” (¶ 521)

As disclosed above, the Court’s
constructions interpret this phrase as “approximating or
roughly calculating values or weights.” Dr. Pazzani
asserts that the scores associated with the categories are
the “values or weights” required by the claim language.
(Pazzani at 238.) But those values are computed via the
equations cited on page 238 in Dr. Pazzani’s report.
4

All citations are to the Fox Report (D.I. 557, Ex. 1). All alterations and emphasis are in the
original.

-6-

(¶ 547)

and the Court
explicitly
rejected
construing
“estimating”
as
“calculating.” (¶ 561)
Rather, all of the alleged parameters are calculated, which
the Court explicitly rejected as a construction for
“estimating.” (¶ 656)
As discussed above, computing a weight is insufficient to
estimate a parameter under the Court’s constructions.
(¶ 661)
Trying to spin the Report, Google cites to Fox’s deposition testimony and attempts to
recast the argument by focusing on the “preciseness” of the number that is the result of the
estimation. Opp. at 9-10. But none of the quotes cited above state that estimating is not satisfied
because the result is too “precise.” Dr. Fox, in fact, did not argue non-infringement based on the
purported “precision” of Google’s estimates.

Instead, he makes the blanket statement—

repeatedly—that calculating/computing cannot be estimating. Period. These opinions are based
on the incorrect premise that “[t]he Court explicitly rejected that ‘estimating’ could be met by a
calculation.” (Fox Report ¶ 290) But the Court made no such finding. The Court simply held
that “estimating” meant “approximating or roughly calculating” and that such construction was
not inconsistent with Bayesian statistics. (D.I. 347, at 32-33). The Court never stated that its
construction could not be satisfied by a calculation, not could it as Bayesian statistical estimates
are typically the result of calculations.
Nor did the Court impose any limitation on the precision of the estimate. “Further, the
extrinsic evidence supplied by PUM supports PUM’s contention that ‘estimating’ was generally
understood by one of ordinary skill in the art at the relevant time as a measurement that is not
entirely precise.” (D.I. 347 at 33 (emphasis added)). Nothing in this statement imposed any

-7-

degree of precision.

But, even if preciseness were a limitation (which it is not), Google

misapprehends what “precision” means.

Google suggests that a rounded number is an

“estimate,” but that a specific number equates to “precision” and thus is not an “estimate.” See,
e.g., Opp. at 9 (stating U.S. population as “‘roughly 300 million’ is estimation, whereas” census
determination that population “is 303,124,754 is not”).5 Under Google’s logic, a mechanic’s
estimate—based on his prior experiences on similar vehicles—that auto repairs will be about
$500 would be an “estimation.” But if the mechanic were to add estimated sales tax of 8.25%,
making for a total estimate of the non-round number of $541.25, it would be a precise
calculation. That is a meaningless cosmetic distinction: either way, the amount is still nothing
more than a prediction.
The example that Google gives to explain this new limitation demonstrates the absurdity
of its argument.

Regardless of the degree of precision, the claimed “estimates” are estimates because
they use observed results to attempt to predict values that by their very nature are not possible to
calculate with 100% accuracy (e.g., degree of user interest in an unseen document). Google’s
attempt to convert the Court’s one sentence observation on extrinsic evidence into a limitation

5

The census number is also an estimate (albeit more precise) because (i) people are born, pass
away, immigrate and emigrate during the process, and (ii) some people are missed during the
counting.

-8-

that the Court did not require and that misunderstands “estimates” as used in Bayesian statistics
should be rejected.
B.

Google Again Raises Irrelevant Issues to Avoid the Real Issue.

Google next tries to turn the table again and suggests that Dr. Pazzani did not apply the
Court’s construction for “estimating.” Opp. at 2, 10. First, that is irrelevant to whether Dr. Fox
followed the Court’s construction. And second, Google is wrong. Dr. Pazzani set forth the
Court’s construction for “estimating,” and stated that this opinions “are based on . . . the Court’s
claim construction.” (Sistos Decl. Ex. 1 ¶¶ 140, 143.)
Google also cites to a reexamination response and alleges that PUM’s Motion is
inconsistent with that response. Contrary to Google’s argument, the reexamination response did
not specifically address whether a computation could satisfy the “estimating” requirement.
Rather, PUM argued that the limitation was not satisfied because the calculation did not include
any estimates. PUM’s infringement theory is entirely consistent with that response. In addition,
the reexaminations are not relevant to the litigation and the response Google cites was filed
before the Court construed the claims.
IV.

THE COURT SHOULD EXCLUDE DR. FOX FROM TESTIFYING THAT
PROBABILITY MUST BE BETWEEN O AND 1.
Google’s Opposition states that “Dr. Fox Will Not Testify That a Probability Must Be

Between 0 and 1.” Opp. at 11. In light of this representation, Google should not have any
objection to the Court entering an order that prohibits Dr. Fox from testifying about such
opinions set forth in his expert report. Since that is the relief that PUM had requested from the
Court, PUM does not address Google’s Opposition directed to “probability” in detail. However,
PUM disagrees with Google’s arguments that (1) PUM’s application of the Court’s construction
is inconsistent with PUM’s validity theories, and (2) Dr. Pazzani pointed to “simply numbers and

-9-

not numerical degrees of belief or likelihood.” Opp. at 2-3, 11-13. Google’s briefing on these
issues contains many misstatements of fact. But since those misstatements are not germane to
this issue, PUM will not further address them.
V.

THERE IS NO BASIS TO EXCLUDE ANY OF DR. CARBONELL’S OPINIONS.
In a footnote at the end of its Opposition, Google asks that “the Court similarly bar

PUM’s expert Dr. Carbonell from raising the same arguments in asserting that the patents are
valid.” Opp. at 14 n.5. Google did not file a motion to exclude any of Dr. Carbonell’s report,
however. Local Rule 7.1.2 requires that “all requests for relief shall be presented to the Court by
motion.”

Further, as explained above, Google did not identify any statements from

Dr. Carbonell’s report that are inconsistent with the Court’s constructions. Thus, there is no
basis to exclude any of his opinions.
VI.

CONCLUSION
The Court already rejected the requirement that “specific” meant “unique,” “exclusive,”

or “his own.” Paragraphs 67, 313, 317-19, 324, 326-27, 330, 336, 384-85, 406, 437, 465, 489,
522, 548, 562, 579, 657, 706, 783, and 791 of the Fox Report rely on the rejected construction
and therefore Dr. Fox should be precluded from offering such opinions at trial. Additionally, the
Court’s claim construction did not say that estimating could not be satisfied by a calculation.
Nor did the Court impose any restrictions on the precision of the estimate. Therefore, Dr. Fox
should be precluded from offering opinions based on paragraphs 288, 290-92, 383, 405, 435,
464, 498, 521, 547, 561, 578, 656, 661, 668, 673, 677, 706, 724, 741, and 785 of his report,
which rely on these imported limitations.

- 10 -

MORRIS, NICHOLS, ARSHT & TUNNELL LLP

/s/ Regina Murphy
_________________________________________
Karen Jacobs (#2881)
Jeremy A. Tigan (#5239)
Regina Murphy (#5648)
1201 N. Market Street
P.O. Box 1347
Wilmington, DE 19899-1347
[phone]
[e-mail]
[e-mail]
[e-mail]
OF COUNSEL:
Marc S. Friedman
DENTONS US LLP
1221 Avenue of the Americas
New York, NY 10020-1089
[phone]

Attorneys for Personalized User Model, L.L.P. and
Yochai Konig

Mark C. Nelson
Steven M. Geiszler
Richard D. Salgado
DENTONS US LLP
2000 McKinney Avenue, Suite 1900
Dallas, TX 75201
[phone]
Jennifer D. Bennett
DENTONS US LLP
1530 Page Mill Road, Suite 200
Palo Alto, CA 94304-1125
[phone]
Matthew P. Larson
DENTONS US LLP
1301 K Street, NW
Suite 600, East Tower
Washington, DC 20005-3364
[phone]
February 19, 2014
8016595.1

- 11 -