`
`
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`
`
`
`
`GOOGLE LLC
`Petitioner
`
`v.
`
`UNILOC 2017 LLC
`Patent Owner
`
`
`
`IPR2020-00755
`PATENT 6,366,908
`
`
`
`PATENT OWNER SUR-REPLY
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`
`TABLE OF CONTENTS
`
`Exhibit List ................................................................................................................ ii 
`I. 
`INTRODUCTION .............................................................................................. 1 
`II.  CLAIM CONSTRUCTION ............................................................................... 1 
`A.  The “keyfact” term ..................................................................................... 1 
`B.  Google failed to address the Board’s preliminary finding that claim
`6 does not invoke 35 U.S.C. § 112, ¶ 6 ...................................................... 7 
`C.  Google acknowledges a refusal to recognize the dependent claims
`recite distinct and additional acts ............................................................... 7 
`III.  PETITIONER FAILS TO PROVE UNPATENTABILITY ............................ 10 
`A.  Google failed to defend the Petition against example deficiencies
`arising from a proper interpretation of the “keyfact” term ...................... 10 
`1.  Google overlooks deficiencies arising from the
`“keyfact” term. .................................................................................. 10 
`2.  Google fails to rebut additional distinctions arising
`from the disavowal limiting the form of the “keyfact”
`term. .................................................................................................. 11 
`3.  Additional deficiencies if Board adopts Google’s
`claim interpretation offered in litigation. .......................................... 16 
`4.  Google mischaracterizes the Response as failing to
`dispute an obvious theory raised in the Petition. .............................. 18 
`B.  Example deficiencies arising from Google’s failure to recognize
`claim 6 recites acts which are distinct from one another and from
`additional acts recited in dependent claims. ............................................. 19 
`C.  Example substantive deficiencies arising from the “keyfact
`retrieving step” of claim 6 ........................................................................ 20 
`IV.  CONCLUSION ................................................................................................ 22 
`
`
`
`i
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`
`EXHIBIT LIST
`
`Exhibit Description
`2001
`Google’s Invalidity Contentions in Uniloc 2017 LLC v. Google LLC,
`No. 2:18-cv-553 (E.D. Tex.), dated August 26, 2019.
`
`2002
`
`Exhibit A-18 to Google’s Invalidity Contentions in Uniloc 2017 LLC v.
`Google LLC, No. 2:18-cv-553 (E.D. Tex.), dated August 26, 2019.
`
`
`
`
`
`ii
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`
`I.
`
`INTRODUCTION
`For the reasons given in Uniloc’s Response (“POR”) and herein, Google fails
`to prove any challenged claim to be unpatentable.
`
`II. CLAIM CONSTRUCTION
`A. The “keyfact” term
`The parties agree the “keyfact” term is coined by the challenged ’908 patent,
`is not a term of art, and hence requires construction here. POR 6-7. As Uniloc
`explained in its Response, “[t]o avoid undue expansion of the ‘keyfact’ term beyond
`the acknowledged definitive scope of the disclosure, the term should be construed for
`purposes of this proceeding to mean ‘a factual extraction of a sentence which
`expresses semantic relation between words in the sentence in the form of [object,
`property].’” POR 9 (quoting Ex. 1001, 4:58-60). It is dispositive here that, under a
`proper construction, an example patentable feature of a “keyfact” is that its format
`itself intrinsically expresses semantic relation between paired words.
`While the intrinsic evidence speaks for itself in support of the above
`construction, certain undisputed observations are worth repeating for emphasis. First,
`the ’908 patent universally qualifies all keyfacts as follows: “All keyfacts express
`semantic relation between words in the form of [object, property].” Ex. 1001, 4:58-
`60. This universal qualification of “all keyfacts” constitutes an objective disavowal
`or disclaimer that limits claim scope. X2Y Attenuators, LLC v. International Trade
`Com’n, 757 F.3d 1358, 1362 (Fed. Cir. 2014) (finding statement that a feature was
`“universal” to “all embodiments” was a clear disavowal that limited claim scope).
`
`
`
`
`
`1
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`Google ignores altogether that the statement in question, on its face, is universally
`applicable to “all keyfacts” in general. Ex. 1001, 4:58-60.
`Second, every single “keyfact” example disclosed in the specification
`expresses semantic relation between words in the paired form of [object, property].
`POR 7-8 (citing, inter alia, Ex. 1001 at Table 1, 1:8-10, 1:16-18, 4:58-60, 6:15-30,
`and 6:38-44). The Abstract emphasizes the paired aspect of a keyfact as follows: “A
`keyfact-based text retrieval method and a keyfact-based text index method that
`describes the formalized concept of a document by a pair comprising an object that
`is the head and a property that is the modifier and uses the information described by
`the pairs as index information for efficient document retrieval.” The consistency with
`which all example “keyfacts” are described further underscores the universal nature
`of the definitive statement, “All keyfacts express semantic relation between words in
`the form of [object, property].” Ex. 1001 at 4:58-60; see also AstraZeneca LP v.
`Apotex, Inc., 633 F.3d 1042, 1052 (Fed. Cir. 2010) (“[W]hen a patentee uses a claim
`term throughout the entire patent specification, in a manner consistent with only a
`single meaning, he has defined that term ‘by implication.’”) (citations omitted).
`Third, as Google acknowledges, Uniloc observed that the Petition takes an
`overly expansive view of “keyfact” that “would impermissibly compass disparaged
`art” cited during prosecution. Reply 2 (quoting POR 6-7). Tellingly, Google does
`not deny these observations concerning either the construction applied in the Petition
`or its inconsistency with the prosecution history.
`
`
`
`
`
`2
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`Rather than defend the construction applied in the Petition in view of the above
`counterpoints raised in Uniloc’s Response, Google offers the conclusory statement
`that it would “contradict[] the intrinsic evidence” to construe “keyfact” as “a factual
`extraction of a sentence which expresses semantic relation between words in the
`sentence in the form of [object, property].” Reply 1. Google fails in its Reply to
`explain why this is allegedly so. Google is wrong for at least the unrebutted reasons
`presented in Uniloc’s Response as to why the intrinsic evidence supports—and
`indeed compels—Uniloc’s proposed construction. POR 6-7.
`Google makes no attempt to explain its conclusory characterization of the
`intrinsic evidence in view of Google’s contrary positions offered in district court.
`Compare Reply 1 with Ex. 1005 at 10.1 In its claim construction briefing in parallel
`litigation, Google acknowledged that “[t]he ’908 patent teaches that keyfacts are
`represented ‘by a pair comprising an object that is the head and a property that is the
`modifier’ in the form of ‘[object, property].’” Ex. 1005 at 10 (citing ’908 patent at
`Abstract, 4:58-60, 6:38-44); see also id. at 17 (arguing “keyfact terms are represented
`in the form ‘[object, property]’”). A construction which simply reflects what Google
`has acknowledged are definitive teachings in the specification cannot reasonably be
`characterized as contradicting the intrinsic evidence.
`Google has not persuasively defended the construction applied in the Petition
`by pointing to the single statement in the specification that “A keyfact means an
`
`
`1 Uniloc cites herein to the pagination in the litigation briefing as it appears in the
`bottom middle of the page, as originally filed in litigation.
`3
`
`
`
`
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`important fact contained in sentences which constitute a document.” Reply 3 (quoting
`’908 patent, 1:15-16). While Uniloc acknowledges here (as it did in litigation) that
`the quoted statement is definitional, it is not the only definitional statement for the
`“keyfact” term provided in the specification. Google offers no basis to dispute that
`the “keyfact” term is further defined and limited by the clarifying and universally
`qualifying statement, “All keyfacts express semantic relation between words in the
`form of [object, property].” Ex. 1001, 4:58-60. In addition, Google offers no legal
`basis to conclude that additional lexicographic statements or disavowal relevant to
`interpreting a claim term can be ignored where a specification also offers a more
`generalized definitive statement for that term.
`Google faults Uniloc for allegedly offering inconsistent claim construction
`positions. Reply 3-4. But it is Google who has offered irreconcilable positions. As
`emphasized above, Google argued in litigation that “[t]he ’908 patent teaches that
`keyfacts are represented ‘by a pair comprising an object that is the head and a property
`that is the modifier’ in the form of ‘[object, property].’” Ex. 1005, 10 (citing ’908
`patent at Abstract, 4:58-60, 6:38-44); see also id. at 17 (arguing “keyfact terms are
`represented in the form ‘[object, property]’”). Indeed, Google not only argued
`“keyfact” requires a particular paired form of [object, property], but also that the
`“object” portion or “first word in a keyfact is always a noun, i.e., keyword.” Ex.
`1005, 10. Google neglects to bring these inconsistent claim interpretations to the
`attention of the Board. Google simply abandons its constructions offered in litigation,
`without even attempting to reconcile them with its contrary positions offered here.
`
`
`
`
`
`4
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`Google also mischaracterizes the statements it partially quotes, out of context,
`from Uniloc’s claim construction briefing in litigation. Reply 5. When read in
`context, the referenced statements simply elaborate why Google not only erred in
`proposing step-plus-function constructions for claim 6, but also in improperly
`incorporating from the specification use of a “keyfact pattern rule” allegedly as a
`necessary (yet unrecited) requirement to perform certain alleged functions. Ex. 1006,
`5-6. Uniloc consistently maintains here that Google erred in construing claim 6 as
`invoking 35 U.S.C. § 112, ¶ 6. The Board preliminarily agreed, finding that “35
`U.S.C. § 112, ¶ 6 does not apply to claim 6, because the claim contains acts to
`perform the cited functions.” Inst. Dec. 17.
`To be clear, no party has asked the Board to construe “keyfact” to require
`nothing more than “facts contained in sentences.” Google never characterizes that
`interpretation as its construction. Indeed, Google offers no construction for the
`“keyfact” term. For purposes of this proceeding, only Uniloc has offered and
`defended a construction for the “keyfact” term—namely, “a factual extraction of a
`sentence which expresses semantic relation between words in the sentence in the form
`of [object, property].” POR 9.
`Google also fails to rebut the alternative arguments Uniloc raised if the Board
`is inclined to interpret “keyfact” differently, and hence in a manner that neither party
`has proposed and defended. For example, even if the Board were to disregard the
`disavowal that “All keyfacts express semantic relation between words in the form of
`[object, property]” (Ex. 1001, 4:58-60), Uniloc nevertheless offered the following
`
`
`
`
`
`5
`
`

`

`
`explanation for why “keyfact” is properly interpreted as an extracted expression of
`fact and not merely a verbatim reproduction of words in sentences:
`
`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`[T]he ’908 patent does not describe or define a “keyfact” as a
`precise semantic representation of the phrase or word pattern of a
`given sentence. That is more akin to the word-based or phrase-
`based approach disparaged in the specification. Ex. 1001, 1:19‒
`28; see also id., 1:42‒45 (“The phrase-based text retrieval
`methods extract a precise phrase pattern though a morphological-
`syntactic normalization process and perform indexing and
`retrieval by extracted phrase.”). In contrast to the disparaged
`approaches,
`the “keyfact”
`term
`itself connotes, and
`the
`surrounding context set forth in the claim language confirms, that
`generation of a keyfact involves a factual derivation extracted
`from keywords.
`POR 8. Uniloc underscored this proper interpretation by observing that, “as shown
`in Table 1, different input sentences which have readily distinguishable semantic
`information may nevertheless produce certain identical keyfact terms.” Id.
`Rather than substantively address the merits of the claim construction
`arguments Uniloc raised in its Response for the “keyfact” term, Google takes the
`position that “the challenged claims are unpatentable over the asserted prior art even
`under” the construction explained in Uniloc’s Response. Reply 3. In view of the
`record before the Board, resolution of the dispute requires determining whether
`Petitioner has met its burden under a construction that correctly reflects all definitive
`statements in the specification—i.e., that “keyfact” means “a factual extraction of a
`sentence which expresses semantic relation between words in the sentence in the form
`of [object, property].” As explained further below, the Board should determine that
`
`
`
`
`
`6
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`Google failed to prove that any challenged claim is unpatentable under this proper
`construction.
`B. Google failed to address the Board’s preliminary finding that
`claim 6 does not invoke 35 U.S.C. § 112, ¶ 6
`Google offers no basis in its Reply for the Board to revisit its determination
`that “35 U.S.C. § 112, ¶ 6 does not apply to claim 6, because the claim contains acts
`to perform the cited functions.” Inst. Dec. 17; POR 9-14. Indeed, Google does not
`even expressly dispute in its Reply that independent claim 6 recites respective acts
`for each recited step. Id. Google simply sidesteps the issue by arguing “the claims
`are unpatentable even under” the Board’s preliminary construction. Reply 5.
`Accordingly, the Board’s preliminary construction for claim 6 stands unrebutted.
`This claim construction issue is itself dispositive as to claim 6, and hence all claims
`6-12 challenged in the Petition. As explained in Uniloc’s Response, and further
`clarified below, “certain example deficiencies of the Petition that arise when the
`challenged claims are properly understood as reciting affirmative acts.” POR 10, 23-
`28.
`
`C. Google acknowledges a refusal to recognize the dependent claims
`recite distinct and additional acts
`Rather than address the Board’s preliminary interpretation that claim 6 recites
`affirmative acts (and thus avoids § 112, ¶ 6), Google disputes whether the Petition
`erred in interpreting the dependent claims as not reciting any “additional acts” with
`respect to those recited in claim 6. Reply 5. Among other reasoning, Uniloc
`explained that it would violate the doctrine of claim differentiation “[t]o suggest any
`
`
`
`
`
`7
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`of the acts recited in claim 6 require nothing more than the additional and expressly
`differentiated acts in claim 7.” POR 15-18.
`Google failed to comprehend, much less rebut, the reasoning Uniloc advanced
`under the doctrine of claim differentiation. Id. As explained in Uniloc’s Response,
`among other limitations, both claim 6 and claim 7 expressly tether respective and
`distinct acts directly to the overarching “keyfact extracting” itself. Id. Claim
`differentiation proscribes conflating into one, for example, the distinct acts of
`“analyzing a document collection and a user query” (recited in claim 6) and
`“analyzing morphology of an input sentence” (recited in claim 7), given both are
`recited as distinct acts for the overarching “keyfact extracting” function itself. Id.
`Google offers no rebuttal to the express differentiation of these distinct acts.
`Google focuses, instead, on claim 8. According to Google, claim 8 does not
`recite additional acts which must be performed, but rather merely defines “a
`previously recited act.” Reply. at 5. Google not only misses the point, but by
`acknowledging it is an “act” that is “previously” recited (id.), Google also essentially
`adopts the Board’s reasoning that “35 U.S.C. § 112, ¶ 6 does not apply to claim 6,
`because the claim contains acts to perform the cited functions” (Inst. Dec. 17).
`Furthermore, Google overlooks claim differentiation expressed in the different ways
`in which dependent claims 7 and 8 introduce their respective limitations. Id.
`While claim 8 directs its limitations to the antecedent act recited in claim 7 as
`“analyzing morphology,” claim 7 recites “analyzing morphology” as an additional
`act required to perform the overarching “keyfact extracting” function of claim 6. This
`
`
`
`
`
`8
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`differentiation reveals that the dependent claims make it explicit as to whether their
`respective limitations recite additional acts for performing an antecedent function
`(e.g., as in claim 7) or, instead, further define an antecedent act (e.g., as in claim 8).
`POR 16 (“In other words, claims 7 and 8 reveal that the dependent claims make
`explicit whether they recite additional acts for the “keyfact extracting step” itself or,
`instead, further define a previously recited act.”).
`A dispositive issue regarding the dependent claims, therefore, is not whether
`the dependent claims recite distinct and additional steps, but rather whether a given
`dependent claim recites distinct and additional acts for performing one of three
`overarching “steps” or functions recited in independent claim 6─namely, (i) “keyfact
`extracting step,” (ii) “keyfact indexing step,” and (iii) “keyfact retrieving step.” That
`certain dependent claims (e.g., claims 7, 10, and 11) recite additional and distinct acts
`only further supports the Board’s preliminary construction that 35 U.S.C. § 112, ¶ 6
`does not apply. Inst. Dec. 17. It also underscores fatal error in that the Petition
`advances no theory that recognizes (1) claim 6 recites acts for performing the “steps”
`set forth therein, and (2) certain dependent claims recite distinct and additional acts
`for performing the “steps” recited in claim 6. POR 15-18.2
`
`
`2 Google chides Uniloc for citing Jenkinson Co. v. Hilton Davis Chem. Co., 520 U.S.
`17, 29 (1997) for the proposition that “[e]ach element contained in a patent claim is
`deemed material to defining the scope of the patented invention.” Reply 9. But then
`Google cites the same opinion for the same proposition. Compare POR 15 with
`Reply 9. Google’s repeated failure to recognize claim limitations as reciting
`affirmative acts falls squarely within the proscriptive warning in Jenkinson and
`violates fundamental claim construction principles. POR 15-18.
`9
`
`
`
`
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`
`III. PETITIONER FAILS TO PROVE UNPATENTABILITY
`The arguments (or lack thereof) in Petitioner’s Reply only underscores
`Petitioner’s failure to prove unpatentability by a preponderance of evidence.
`
`A. Google failed to defend the Petition against example deficiencies
`arising from a proper interpretation of the “keyfact” term
`1. Google overlooks deficiencies arising from the “keyfact” term.
`Google fails to rebut example deficiencies of the Petition that arise from a
`proper construction of the “keyfact” term. Certain example deficiencies arise
`regardless of whether the Board agrees that the intrinsic evidence contains disavowal
`which limits the term to the paired form of [object, property]. As explained above in
`addressing claim construction, the ’908 patent disparages and distinguishes
`approaches which apply verbatim word-based or phrase-based text retrieval. See
`supra, § II.A (quoting POR 8); POR 21-22. “In contrast to the disparaged
`approaches, the ‘keyfact’ term itself connotes, and the surrounding context set forth
`in the claim language confirms, that generation of a keyfact involves a factual
`derivation extracted from keywords.” POR 8.
`Under
`this proper understanding, Uniloc explained certain example
`deficiencies over Braden-Harder as follows:
`
`Using so-called “triples” to symbolically represent the “semantic
`information” of an input string, quite literally verbatim, is
`distinguishable on its face at least from “keyfact” extraction as
`disclosed and claimed. As discussed above in addressing the
`proper construction of the “keyfact” term, the ’908 patent
`disparages the sort of phrase-based scheme Braden-Harder
`employs as being subject to certain technical disadvantages. Ex.
`1001, 1:19‒28; see also id., 1:42‒45.
`
`
`
`
`
`10
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`POR 22. Uniloc offered the above distinctions in addition to Google’s failure to
`establish that Braden-Harder’s “triples” have the required paired form of [object,
`property]. Id. (“Moreover, Braden-Harder’s use of ‘triples’ in its phrase-based
`scheme is distinguishable from claim language directed to extracting keyfacts in the
`paired form of [object, property].”).
`Google overlooks the specific distinction that the verbatim representation of a
`sentence in Braden-Harder is “more akin to the word-based or phrase-based approach
`disparaged in the specification” and lacks “a factual derivation extracted from
`keywords.” See supra, § II.A (quoting POR 8); POR 21-22. Google erroneously
`attempts, instead, to conflate independent patentable distinctions arising from (1) the
`factual derivation aspect of a keyfact and (2) the required form of a keyfact. Reply
`18 (“Uniloc is wrong that Braden-Harder’s triples cannot be a keyfact because the
`triple form ‘literally verbatim’ contains more than the pair of node words.”). While
`both keyfact features present patentable, they are not one and the same. The
`patentable distinctions arising from the factual derivation feature remains unrebutted
`and is independently fatal to the Petition.
`
`2. Google fails to rebut additional distinctions arising from the
`disavowal limiting the form of the “keyfact” term.
`Google fails to rebut the additional fatal deficiency of Braden-Harder arising
`under a proper construction that the “keyfact” term is limited to the paired form of
`[object, property] in expressing semantic relation between words. Attacking only the
`form aspect of the construction, Google falsely asserts that Uniloc “identified nothing
`in the specification or the prosecution history” that supports limiting “keyfact” in
`
`
`
`
`
`11
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`terms of the paired form of [object, property]. Reply 18-19. Google ignores, and
`hence offers no rebuttal to, the intrinsic evidence addressed above and in Uniloc’s
`Response. See supra, § II.A; POR 1-4 and 6-9. This includes, for example, the
`disavowal expressed in the statement “All keyfacts express semantic relation between
`words in the form of [object, property].” Ex. 1001, 4:58-60; see also id. at Abstract,
`1:8-10, 1:16-18, 4:58-60, 6:15-30, 6:38-44.
`Google also points, without explanation, to the table entries of “[KEY2 KEY1,
`NIL]” and “[KEY1 KEY2 KEY3]” (disclosed in Tables 1 and 3, respectively, of the
`’908 patent) as allegedly supporting an interpretation that Braden-Harder’s “triples”
`satisfy, in analogous form, the extracted keyfact terms as disclosed and claimed.
`Reply 19. Google confuses distinct concepts of the ’908 patent.
`The ’908 patent makes an express distinction between keyfact patterns and a
`list of keyfact terms that are extracted using keyfact patterns. With reference to Table
`1, for example, when the input instance “‘the fast retrieval of the distributed
`information’ is applied to the key fact pattern rule[,] . . . the key fact pattern ‘MP1
`KEY1 PO MP2 KEY2’ is the result.” Ex. 1001, 6:34-37. This example keyfact
`pattern is not described as being a keyfact itself. Rather, “[k]eyfact terms that have
`forms of [object, property] are generated as to the input keyfact pattern at the stage
`of the keyfact generation 35 by searching the keyfact generation rule 38.” Id. at 6:38-
`41; see also id. at 3:33-35 (“The fourth step is to apply the keyfact pattern to a keyfact
`pattern generation rule and generating a keyfact list.”)
`
`
`
`
`
`12
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`Google’s unexplained citation to “[KEY1 KEY2 KEY3]” overlooks that
`Table 3 identifies this table entry as being a keyfact pattern, as opposed to a keyfact
`extracted from such a pattern and having the form of [object, property].
`
`
`’908 patent, 7:46-55 (Table 3) (red highlighting added). Accordingly, Google only
`highlights example deficiencies of the Petition by attempting to analogize Braden-
`Harder’s “triples” to something the ’908 patent distinguishes from its keyfact term
`having the form [object, property].
`Google’s unexplained citation to the entry “[KEY2 KEY1, NIL]” of Table 1
`of the ’908 patent fares no better. Reply 19. In that example, the “property” portion
`is “NIL” or empty, such that the KEY2 and KEY1 pair are the only two nonempty
`keyfact tags. This is true for all keyfacts shown in Table 1. In every instance, the
`example keyfacts listed never consist of more than a pair of nonempty tags. See
`supra, § II.A (discussing the paired aspect of keyfacts). This consistent and repeated
`threshold of no more than a pair of nonempty tags is distinguishable on its face from
`Braden-Harder’s nonempty, three-part “triples,” even if one were to ignore that
`
`
`
`
`
`13
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`Braden-Harder’s “triples” are not extracted keyfacts as disclosed and claimed (much
`less in the paired form of [object, property]).
`Contrary to what Google suggests, Uniloc disputes that Braden-Harder’s
`“triples” use an analogous format to exhibit semantic relationship between an object
`and its modifier. Cf. Reply 19. Braden-Harder speaks for itself in refuting such a
`characterization. Near the outset of the disclosure, Braden-Harder describes its
`“triples” as requiring the “form of ‘word-relation-word’” in order to operate as
`intended. Ex. 1020, 5:16-25. More specifically, the “triples” format applied in
`Braden-Harder expressly requires three parts: (1) a first word and (2) a labeled
`“relation” which by its content interlinks the first word with (3) a second word. Id.;
`see also id. at 14:3-4 (“Each triple contains two node words as depicted in the graph
`linked by a semantic relationship therebetween.”).
`Braden-Harder’s labeled “relation” is essential in that its content defines how
`the first and second words are related. Id. By way of contrast, an example patentable
`feature of a “keyfact,” as disclosed and claimed in the ’908 patent, is that its format
`itself intrinsically expresses semantic relation between paired words, thereby
`obviating the need, for example, to expressly label a semantic relation through
`inclusion of a third distinct component.
`Google also falsely suggests it is undisputed that it would have been obvious
`to modify Braden-Harder to exclude from its “triples” the interposing relation label.
`Reply 19-20 (“According to Dr. Jansen, ‘a person of ordinary skill in the art would
`be motivated to generate logic forms that do not include the relations … to increase
`
`
`
`
`
`14
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`matches with ‘sufficiently similar semantic content’ (25:51-52) … [and] to ensure
`that all relevant documents are retrieved from the document collection.”) (citing Ex.
`1003, ¶221 and Pet. 47). Even if one were to ignore that such a modification would
`render Braden-Harder inoperable for its intended purpose, it would only further
`distinguish Braden-Harder from claim language directed to the “keyfact” term. As
`discussed above, and in Uniloc’s Response, the “keyfact” term is properly construed
`to mean “a factual extraction of a sentence which expresses semantic relation between
`words in the sentence in the form of [object, property].’” See § II.A, supra (citing
`POR 9, which quotes Ex. 1001, 4:58-60). This clearly cannot be shown obvious by
`removing the only aspect of Braden-Harder’s “triples” that Google asserts as
`allegedly satisfying the “semantic relation” requirement.
` Google also fails to defend its proposed modification of Braden-Harder (by
`removing the interposing relation label) as not rendering the disclosure inoperable for
`its intended purpose. In re Gordon, 733 F.2d 900, 902 (Fed. Cir. 1984) (if proposed
`modification would render the prior art invention being modified unsatisfactory for
`its intended purpose, then there is no suggestion or motivation to make the proposed
`modification). Braden-Harder emphasizes the importance of the interposing relation
`label of its “triples” throughout its disclosure. See, e.g., Ex. 1020, Abstract (“Each
`such logical form encodes, in a word-relation-word manner, semantic relationships,
`particularly argument and adjunct structure, between words in a phrase.”); 11:41-43
`(“Broadly speaking, a logical form is a directed acyclic graph in which words
`representing text of any arbitrary size are linked by labeled relations.”); 13:29-31
`
`
`
`
`
`15
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`(“To identify all the semantic relationships in an input string, each node in the
`syntactic parse tree for that string is examined.”); 14:15-17 (describing “three logical
`form constructions for which additional natural language processing is required to
`correctly yield all the logical form triples”); 23:29-31 (“all remaining document
`records are assigned a score as defined above and based on the relation type(s) of
`matching triples”). As shown by the above non-exhaustive examples, Braden-Harder
`speaks for itself in refuting the conclusory theory that it would have been obvious to
`remove the only disclosed aspect of “triples” which conveys “relation.”
`
`3.
`
`Additional deficiencies if Board adopts Google’s claim
`interpretation offered in litigation.
`Google suggests in its Reply that the “keyfact” term is indiscriminate as to the
`object and property portions. Reply 15. In making its point, Google takes umbrage
`with Uniloc citing to disclosure in the ’908 patent which “clarifies that ‘the class of
`noun determines the object or the property of the keyfacts.’” Id.; POR 21 (citing Ex.
`1001, 5:35-36).
`Contrary to what Google suggests, Uniloc did not cite the above passage to
`argue certain classes of nouns must always be considered a property, while others
`must always be considered an object. Reply 15. Rather, Uniloc observed, and Google
`failed to dispute, that the form of [object, property] itself conveys semantic relation.
`In applying this form, different information may be intrinsically conveyed, for
`example, when a keyfact tag is represented as an object as opposed to a property, or
`when tags are listed in a particular order. POR 21-22.
`
`
`
`
`
`16
`
`

`

`IPR2020-00755
`U.S. Patent No. 6,366,908
`
`
`The intrinsic evidence supports this understanding. See § II.A, supra. For
`example, the ’908 patent teaches keyfact generation may involve determining
`whether a given keyfact tag is recorded as an object or as a property, or determining
`the order in which such tags are listed. Id. The ’908 patent also teaches keyfacts may
`convey different information depending on whether a given tag is listed as an object
`as opposed to a property. For example, Table 1 separately lists both “[KEY2,
`KEY1]” and “[KEY2 KEY1, NIL]” as distinct keyfact terms extracted from the
`statement “the retrieval of information.” Ex. 1001, 6:16-18. While the same
`nonempty tags KEY2 and KEY1 are used in these two examples, in one instance both
`tags are represented as objects (“[KEY2 KEY1, NIL]”), and in the other one is
`represented as an object and the other is represented as a property (“[KEY2, KEY1]”).
`It would be unnecessarily redundant to list both keyfacts separately if each conveyed
`the exact same information. That both are included in the same list confirms t