`——————————
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`——————————
`GOOGLE LLC,
`
`Petitioner,
`
`v.
`
`UNILOC 2017 LLC,
`
`Patent Owner.
`——————————
`Case No. IPR2020-00755
`U.S. Patent No. 6,366,908
`Filing Date: December 30, 1999
`Issue Date: April 2, 2002
`
`
`
`PETITIONER’S REPLY TO PATENT OWNER’S RESPONSE
`
`
`
`
`
`
`
`TABLE OF CONTENTS
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`B.
`
`C.
`
`D.
`
`INTRODUCTION ........................................................................................... 1
`I.
`II. UNILOC’S CLAIM CONSTRUCTION ARGUMENTS FAIL ..................... 2
`A.
`The Board need not address Uniloc’s new construction of the
`term “keyfact.” ...................................................................................... 2
`To the extent the Board wishes to construe the term “keyfact,”
`it should reject Uniloc’s new construction because that
`construction, POR 6-9, cannot be squared with the intrinsic
`record and contradicts Uniloc’s litigation position. .............................. 3
`The Board should not accept Uniloc’s Section 112, ¶6
`arguments, POR 9-14, which improperly change the meaning of
`the claims. .............................................................................................. 4
`The Board should reject Uniloc’s misapplication of “the
`doctrine of claim differentiation and the ‘all elements’ rule” to
`require certain dependent claims to recite “additional and
`expressly differentiated acts,” POR 14-19. ........................................... 5
`1.
`The doctrine of claim differentiation and the “all
`elements” rule do not support Uniloc’s inconsistent
`additional claim requirements. .................................................... 5
`III. THE CHALLENGED CLAIMS ARE UNPATENTABLE ............................ 9
`A.
`Braden-Harder renders obvious the claimed “keyfact” under
`both Uniloc’s litigation and new constructions. .................................... 9
`1.
`Uniloc does not dispute that Braden-Harder renders
`obvious the claimed “keyfact” under Uniloc’s litigation
`construction. ................................................................................ 9
`Braden-Harder renders obvious the claimed “keyfact”
`under Uniloc’s new construction. ............................................. 10
`Uniloc’s alleged “deficiencies” related to its new
`construction, POR 20-22, do not impact the outcome of
`this case. .................................................................................... 13
`
`2.
`
`3.
`
`i
`
`
`
`
`
`a.
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`The Petition and evidence show that Braden-
`Harder’s logical form triples are “a factual
`extraction of a sentence.” ................................................ 13
`The claims do not require “determination between
`a class of noun.” .............................................................. 15
`Braden-Harder determines between a class of
`noun. ............................................................................... 16
`Braden-Harder’s logical form triples represent “a
`paired relationship” between an object and a
`property. .......................................................................... 18
`B. Uniloc’s rehashes of its POPR arguments do not call into
`question how Braden-Harder renders the claims obvious.................. 20
`1.
`Uniloc’s arguments that Braden-Harder is a phrase-
`based text retrieval method, POR 8, 22, fail. ............................ 20
`Uniloc’s arguments regarding the relationship between
`claim 6 and the dependent claims, POR 23-24, fail. ................ 24
`Uniloc’s arguments regarding example deficiencies
`arising from the keyfact retrieving step, POR 24-28, fail......... 27
`C. Uniloc does not dispute that the dependent claims add nothing
`patentable. ............................................................................................ 30
`IV. CONCLUSION .............................................................................................. 31
`
`
`b.
`
`c.
`
`d.
`
`2.
`
`3.
`
`
`
`
`
`ii
`
`
`
`
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`TABLE OF AUTHORITIES
`
`Federal Cases
`Auburn Univ. v. Int’l Bus. Machines, Corp.,
`No. 3:09-CV-694-MEF, 2011 WL 13225006 (M.D. Ala. Apr. 22, 2011) ........... 6
`In re Neurografix (’360) Patent Litig.,
`201 F. Supp. 3d 206 (D. Mass. 2016) ................................................................... 4
`In re Skvorecz,
`580 F.3d 1262 (Fed. Cir. 2009) .......................................................................... 19
`Kraft Foods, Inc. v. Int’l Trading Co.,
`203 F.3d 1362 (Fed. Cir. 2000) ........................................................................7, 8
`Masco Corp. v. United States,
`303 F.3d 1316 (Fed. Cir. 2002) ............................................................................ 4
`Phillips v. AWH Corp.,
`415 F.3d 1303 (Fed. Cir. 2005) ............................................................................ 3
`Realtime Data, LLC v. Iancu,
`912 F.3d 1368 (Fed. Cir. 2019) .................................................................. passim
`Saffran v. Johnson & Johnson,
`712 F.3d 549 (Fed. Cir. 2013) .............................................................................. 7
`Seal-Flex, Inc. v. Athletic Track and Court Const.,
`172 F.3d 836 (Fed. Cir. 1999) .............................................................................. 4
`Vivid Techs., Inc. v. Am. Sci. & Eng’g, Inc.,
`200 F.3d 795 (Fed. Cir. 1999) .............................................................................. 3
`Warner-Jenkinson Co. v. Hilton Davis Chem. Co.,
`520 U.S. 17 (1997) ................................................................................................ 9
`
`iii
`
`
`
`
`I.
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`INTRODUCTION
`Abandoning the construction it advanced in litigation for the term “keyfact,”
`
`Uniloc’s Patent Owner Response (POR) introduces a new construction. POR 9.
`
`Uniloc now seeks to narrowly define “keyfact” as “a factual extraction of a
`
`sentence which expresses semantic relation between words in the sentence in the
`
`form of [object, property].” POR 9. This new construction contradicts the intrinsic
`
`evidence. Regardless, this new construction does not impact the outcome of this
`
`proceeding.
`
`If the Board rejects Uniloc’s new construction, Uniloc does not dispute that
`
`Braden-Harder’s extracted logical form triples are facts contained in sentences,
`
`meet the keyfact limitation, and provide a concept-based retrieval system. Pet. 34;
`
`Jansen Decl., ¶176 (“Ex. 1003”). Alternatively, even if the Board adopts Uniloc’s
`
`new construction, as Google and its expert, Dr. Jansen, previously explained,
`
`Braden-Harder’s logic form triples are factual extractions of a sentence that
`
`express the semantic relation between words in the sentence. Pet. 45-47; Ex. 1003
`
`¶¶211-219. Moreover, for the case of noun modifiers, these logic form triples are
`
`expressed in the same form as Uniloc’s new, narrower construction of keyfact. See
`
`Pet. 47, and Ex. 1003 ¶¶221-222. Not only does Uniloc fail to provide any contrary
`
`evidence or expert testimony, but Uniloc also fails to address, let alone dispute, the
`
`relevant portions of Google’s Petition and expert testimony. Thus, the challenged
`
`1
`
`
`
`
`claims are unpatentable over the asserted prior art even under Uniloc’s new
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`construction.
`
`With no supporting expert testimony or other evidence, Uniloc also uses its
`
`POR to rehash arguments from its Patent Owner Preliminary Response (POPR).
`
`Compare POR 14-19, 23-24, with POPR 19-23; compare POR 8, 22, with POPR
`
`25-27; compare POR 24-28, with POPR 27-31. But these arguments fail for the
`
`same reasons that the Board identified when rejecting them in its Institution
`
`Decision and for the additional reasons discussed below. Institution Decision 17-
`
`18, 30-31, 33-34.
`
`II. UNILOC’S CLAIM CONSTRUCTION ARGUMENTS FAIL
`A. The Board need not address Uniloc’s new construction of the term
`“keyfact.”
`The Petition proposed that the Board use Uniloc’s litigation construction for
`
`the term “keyfact”: “fact contained in sentences.” Pet. 24. In the Institution
`
`Decision, the Board agreed that Braden-Harder’s logical form triples are
`
`“keyfacts” under Uniloc’s litigation construction. Institution Decision 30-31.
`
`Uniloc now asserts that its own construction “unduly broaden[s] [the] claim
`
`scope beyond how the ‘keyfact’ term is consistently used in the ’908 patent and in
`
`a manner that would impermissibly encompass disparaged art,” POR 6-7, and
`
`seeks to limit the term with the following additions as shown in bold/italics below:
`
`2
`
`
`
`
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`a factual extraction of a sentence which expresses semantic
`relation between words in the sentence in the form of [object,
`property].
`
`POR 9 (emphases added). However, the Board need not construe the term
`
`“keyfact” here because it is not necessary to resolve the controversy. Vivid Techs.,
`
`Inc. v. Am. Sci. & Eng’g, Inc., 200 F.3d 795, 803 (Fed. Cir. 1999). As explained in
`
`the Petition and below, the challenged claims are unpatentable over the asserted
`
`prior art even under Uniloc’s new construction. Pet. 57; supra Section III.
`
`B. To the extent the Board wishes to construe the term “keyfact,” it
`should reject Uniloc’s new construction because that construction,
`POR 6-9, cannot be squared with the intrinsic record and
`contradicts Uniloc’s litigation position.
`Uniloc’s new construction is unduly narrow because it requires keyfacts to
`
`be “expresse[d] … in the form of [object, property]” despite the specification not
`
`defining the term that way. POR 9. According to the Federal Circuit, if “the
`
`specification ... reveal[s] a special definition given to a claim term by the patentee
`
`... the inventor’s lexicography governs.” Phillips v. AWH Corp., 415 F.3d 1303,
`
`1316 (Fed. Cir. 2005) (emphasis added). Here, the specification defines “keyfact”:
`
`“A keyfact means an important fact contained in sentences which constitute a
`
`document.” ’908 patent, 1:15-16.
`
`In the litigation, Uniloc pointed to this same language in the specification
`
`and agreed it was definitional. Ex. 1004 at 6 (“what a keyfact is”). Moreover, in the
`
`3
`
`
`
`
`litigation, Uniloc argued that the term “keyfact” does not require any particular
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`form, stating:
`
` “there is no function recited in claim 6 that would require … keyfacts in a
`
`particular format”; and
`
` Google “improperly” contended that “keyfact terms are represented in the
`
`form ‘[object, property].’”
`
`Ex. 1006 at 6. Accordingly, should the Board decide to construe “keyfact,” the
`
`Board should reject Uniloc’s new construction that requires “keyfact” to have a
`
`particular form.
`
`C. The Board should not accept Uniloc’s Section 112, ¶6 arguments,
`POR 9-14, which improperly change the meaning of the claims.
`As explained in the Petition, claim 6 recites limitations using the “step for”
`
`language, and there is a presumption that these limitations are step-plus-function
`
`limitations governed by Section 112, ¶6. See Masco Corp. v. United States, 303
`
`F.3d 1316, 1326 (Fed. Cir. 2002); Seal-Flex, Inc. v. Athletic Track and Court
`
`Const., 172 F.3d 836, 851 (Fed. Cir. 1999). The Board should decline Uniloc’s
`
`invitation, POR 12-13, to ignore this presumption by rewriting the words “as [step]
`
`‘comprising’ without [allegedly] changing the meaning.” See In re Neurografix
`
`(’360) Patent Litig., 201 F. Supp. 3d 206, 213-14 (D. Mass. 2016) (noting the
`
`significance of the signaling phrases in identifying “step-plus-function terms”)
`
`4
`
`
`
`
`(citation omitted). However, there is no need for the Board to address the issue
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`because the claims are unpatentable even under Uniloc’s argument. See Institution
`
`Decision 17-18.
`
`D. The Board should reject Uniloc’s misapplication of “the doctrine of
`claim differentiation and the ‘all elements’ rule” to require certain
`dependent claims to recite “additional and expressly differentiated
`acts,” POR 14-19.
`1. The doctrine of claim differentiation and the “all elements”
`rule do not support Uniloc’s inconsistent additional claim
`requirements.
`Neither the doctrine of claim differentiation nor the “all elements” rule
`
`supports Uniloc’s proposal that the Board should construe dependent claims 7, 10,
`
`and 11 as “recit[ing] additional acts” (i.e., further adding new “distinct and
`
`additional” steps to claim 6), POR 15, 18-19, but construe claim 8 as “further
`
`defin[ing] a previously recited act,” id., 15-16. First, the language of the claims
`
`does not differ in a way that could support such different constructions. These
`
`claims all identically recite “wherein said step of … comprises the steps of …”
`
`language. ’908 patent, 9:54-10:57. As such, these claims should be properly
`
`construed as describing details required as part of performing the previously
`
`introduced steps. See, e.g., Realtime Data, LLC v. Iancu, 912 F.3d 1368, 1375
`
`(Fed. Cir. 2019) (“[T]he claim elements introduced in dependent claim 4 with
`
`‘comprising’ language are properly understood as giving details sufficient to
`
`5
`
`
`
`
`constitute a particular embodiment of the more general ‘maintaining a dictionary’
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`term of independent claim 1.”); Auburn Univ. v. Int’l Bus. Machines, Corp., No.
`
`3:09-CV-694-MEF, 2011 WL 13225006, at *4 (M.D. Ala. Apr. 22, 2011) (noting
`
`claims describing “limitations on how the [previously recited] step is to be done”
`
`by using the “where the … step comprises” language while others “add additional
`
`steps” by using the “where the … step further comprises” language (emphasis
`
`added)).
`
`Realtime Data is instructive. In Realtime Data, the Federal Circuit reviewed
`
`dependent claim 4, which recites “wherein the step of maintaining a dictionary
`
`comprises the steps of …” language (similar to the dependent claims of the ’908
`
`patent). Realtime Data, 912 F.3d at 1375-76. According to the Federal Circuit, the
`
`“comprising” language in claim 4 does not add additional steps and, instead, gives
`
`“details sufficient to constitute a particular embodiment of the more general” term
`
`of the independent claim. Id. at 1375. Thus, “the steps outlined in dependent claim
`
`4 were sufficient to satisfy the ‘maintaining a dictionary’ limitation in independent
`
`claim 1.” Id. at 1375. Here, the ’908 patent’s similar use of “wherein said step of
`
`… comprises the steps of …” in dependent claims 7, 10, and 11 gives details to the
`
`previously claimed steps in claim 6; it does not add new “distinct and additional”
`
`steps—contrary to Uniloc’s contentions. POR 16, 19.
`
`6
`
`
`
`The specification further confirms that these dependent claims give details
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`
`
`of the more general previously recited steps; they do not add new “distinct and
`
`additional” steps. ’908 patent, 3:27-67. The specification first summarizes the
`
`purpose of each of the “keyfact extracting step, keyfact indexing step, and keyfact
`
`retrieving step.” Id., 3:11-13; see also id., 4:22-28, Figs. 1-2. Then the
`
`specification goes on to describe what each of the steps includes (i.e., the way each
`
`step is to be accomplished). Id., 3:27-36 (“The step of keyfact extracting includes
`
`the following steps.”), 3:49-55 (“The step of keyfact indexing includes the
`
`following steps.”), 3:56-67 (“The step of keyfact retrieving includes the following
`
`steps.”).
`
`Therefore, the dependent claims should be interpreted as giving details of
`
`the previously recited more general steps, not adding new “distinct and additional”
`
`steps. Realtime Data, 912 F.3d at 1375. Uniloc’s reliance on the doctrine of claim
`
`differentiation is inapposite. It merely creates “a presumption that each claim in a
`
`patent has a different scope;[1] it is ‘not a hard and fast rule of construction.’” Kraft
`
`Foods, Inc. v. Int’l Trading Co., 203 F.3d 1362, 1368 (Fed. Cir. 2000). “Moreover,
`
`
`1 However, claim differentiation cannot broaden claims beyond the corresponding
`
`structure specifically disclosed in the specification under Section 112, ¶6. See, e.g.,
`
`Saffran v. Johnson & Johnson, 712 F.3d 549, 563 (Fed. Cir. 2013).
`
`7
`
`
`
`
`that the claims are presumed to differ in scope does not mean that every limitation
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`must be distinguished from its counterpart in another claim, but only that at least
`
`one limitation must differ.” Id. at 1368. Uniloc references the doctrine for the
`
`unremarkable principle that “[t]he presence of a dependent claim that adds a
`
`particular limitation gives rise to a presumption that the limitation in question is
`
`not present in the independent claim.” POR 15 (citation omitted). But that principle
`
`is not applicable here, as Google submits that the dependent claims do further limit
`
`the independent claims. See Realtime Data, 912 F.3d at 1375-76.
`
`Uniloc is thus wrong that Google “erroneously conflat[ed]” the “keyfact
`
`extracting step,” “the ‘keyfact indexing step’ and the ‘keyfact retrieving step’ of
`
`claim 6 with other distinct and additional acts recited in dependent claims.” POR
`
`17-19. Rather than “impermissibly read[ing] out limitations,” as Uniloc contends,
`
`the Petition provides the most reasonable and consistent read of the claims in view
`
`of the specification:
`
`1. Should Section 112, ¶6 apply, the acts of the step-for limitations come
`
`from the steps described in the specification.
`
`2. Should Section 112, ¶6 not apply, the dependent claims merely recite
`
`steps that give details of the more general steps of claim 6.
`
`8
`
`
`
`
`The challenged claims are unpatentable either way.2 See Institution Decision 17-
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`18.
`
`Finally, Uniloc references the “all elements” rule, but that rule, articulated in
`
`Warner-Jenkinson, is likewise not applicable to the unpatentability analysis here.
`
`Instead, the rule relates to the doctrine of equivalents for infringement, specifically
`
`the requirement to show equivalence for all individual elements of the claims
`
`because “[e]ach element contained in a patent claim is deemed material to defining
`
`the scope of the patented invention.” Warner-Jenkinson Co. v. Hilton Davis Chem.
`
`Co., 520 U.S. 17, 29 (1997).
`
`III. THE CHALLENGED CLAIMS ARE UNPATENTABLE
`A. Braden-Harder renders obvious the claimed “keyfact” under both
`Uniloc’s litigation and new constructions.
`1. Uniloc does not dispute that Braden-Harder renders obvious
`the claimed “keyfact” under Uniloc’s litigation construction.
`The asserted “prior art teaches the disputed claim terms … in the same
`
`manner as the ’908 patent,” and “[Braden-Harder’s] logical-form triples,
`
`
`2 Should Section 112, ¶6 apply, the asserted art performs each of the claimed
`
`functions using the same or equivalent structures and acts as the ’908 patent, and
`
`thus the challenged claims are unpatentable under Section 112, ¶6 for the reasons
`
`discussed in the Petition. Pet. 47; Ex. 1003, ¶223.
`
`9
`
`
`
`
`equivalent to the claimed keyfacts, are facts contained in sentences (Uniloc’s
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`[litigation] construction of keyfacts).” Pet. 24, 34; Ex. 1003 ¶176. Thus, Braden-
`
`Harder renders obvious the claimed “keyfact” under Uniloc’s litigation
`
`construction. Pet. 24, 34; Institution Decision, 28, 31, 34. Uniloc does not dispute
`
`this in the POR. POR 20.
`
`2. Braden-Harder renders obvious the claimed “keyfact” under
`Uniloc’s new construction.
`Under Uniloc’s new construction, see supra Section II.A, the challenged
`
`claims remain unpatentable over the asserted prior art because—as explained in the
`
`Petition—Braden-Harder’s extracted logical forms are each “a factual extraction
`
`of a sentence which expresses semantic relation between words in the sentence in
`
`the form of [object, property].” POR 9; see Pet. 47; Ex. 1003 ¶¶221-222.
`
`As the Petition explains, Braden-Harder teaches various special graph
`
`walks, including one for “the case of a constituent with noun modifiers.” Pet. 45
`
`(quoting Braden-Harder, 14:37-41, Fig. 5D); Ex. 1003 ¶¶216-217.
`
`10
`
`
`
`
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`
`
`Braden-Harder, Fig. 5D (annotated); Ex. 1003 ¶216; Pet. 46.
`
`The Petition next explains that should keyfacts “have [the] forms of [object,
`
`property] …, a POSITA would have understood that Braden-Harder generates
`
`logical form triples for ‘the case of a constituent with noun modifiers’ (see Table
`
`3), such as ‘Nadj’ and ‘Mods’ relations, which exhibit the semantic relationship of
`
`the form [object, modifier].” Pet. 47; Ex. 1003 ¶¶221. Braden-Harder’s Table 3,
`
`reproduced below, shows a subset of the possible semantic relations:
`
`
`
`11
`
`
`
`
`Braden-Harder, Table 3 (annotated); see also Pet. 58; Ex. 1003 ¶264. As shown in
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`Braden-Harder, the Nadj relation (highlighted green above) expresses the
`
`semantic relation of an “adjective modifying a noun” and the Mods relation
`
`(highlighted blue above) expresses another “modifier that is not a clause” that
`
`could modify a noun. Thus, in the case of a “noun modifier,” including at least the
`
`Nadj and Mods relations, Braden-Harder’s logical form triple expresses the
`
`semantic relationship of a noun (e.g., object) and its corresponding modifier (e.g.,
`
`property), see Pet. 47; Ex. 1003 ¶¶221, and thus is “a pair comprising an object that
`
`is the head and a property that is the modifier,” ’908 patent, 1:7-12. See also Pet.
`
`34 (Each logical-form triple uses a pair of node words linked by “a semantic
`
`relationship therebetween” and “convey[s] the semantic information inherent in
`
`input string….”); Ex. 1003 ¶176.
`
`Uniloc neither disputes the relevant discussion in the Petition nor challenges
`
`Dr. Jansen’s explanation. See POR 20-22; see generally, Pet. 47; Ex. 1003 ¶¶221-
`
`222. Indeed, the POR’s discussion on page 22 of “what [Braden-Harder’s] ‘triple’
`
`form represents” completely fails to address Braden-Harder’s logical form triples
`
`for the case of a constituent with noun modifiers. Accordingly, for the reasons
`
`discussed in the Petition, Pet. 47, explained by Dr. Jansen, Ex. 1003 ¶¶221-222,
`
`and ignored by Uniloc, the claimed “keyfact” would be obvious in view of the
`
`asserted prior art because Braden-Harder’s extracted logical form is “a factual
`
`12
`
`
`
`
`extraction of a sentence which expresses semantic relation between words in the
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`sentence in the form of [object, property],” even under Uniloc’s new construction.
`
`3. Uniloc’s alleged “deficiencies” related to its new construction,
`POR 20-22, do not impact the outcome of this case.
`a. The Petition and evidence show that Braden-Harder’s
`logical form triples are “a factual extraction of a
`sentence.”
`Uniloc contends that Braden-Harder “is distinguishable on its face from the
`
`distinct concept of a ‘keyfact’ that is an extracted expression of fact.” POR 21
`
`(emphasis in original). But, again, Uniloc fails to address the relevant portion of
`
`the Petition.
`
`On pages 43-47, the Petition explains how the prior art teaches function
`
`[6a3] “respectively extracting keyfacts of said document collection and said user
`
`query from said keywords.” Pet. 43-47 (emphasis added); Ex. 1003 ¶¶209-223. As
`
`part of this explanation, the Petition articulates how Braden-Harder’s “logical
`
`form triples are extracted based on the semantic relationships of the nouns (i.e.,
`
`keywords) with other words in the sentence.” Pet. 45; Ex. 1003 ¶211.
`
`As Dr. Jansen explained, Braden-Harder uses generation rules to generate a
`
`list of logical form triples. Ex. 1003 ¶211; Pet. 45. For example, “for the input
`
`string, ‘I like shark fin soup bowls,’ a list of eight logical form triples is
`
`generated,” as shown below. Ex. 1003 ¶217; Pet. 45.
`
`13
`
`
`
`
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`
`
`Braden-Harder, Fig. 5D (annotated); Ex. 1003 ¶216; Pet. 46.
`
`Because the logical form triples are extracted based on the semantic
`
`relationships of the nouns, each of Braden-Harder’s logical form triples is “a
`
`factual extraction of a sentence which expresses semantic relation between words
`
`in the sentence.”
`
`The Board recognized this when it previously rejected similar arguments
`
`raised in the POPR. Institution Decision 31 (“Patent Owner’s arguments do not
`
`consider Petitioner’s showing that … [the] list of logical form triples … are
`
`extracted based on semantic relationships of the nouns (keywords) with other
`
`words in the sentence.”). Moreover, for the noun modifier cases, those logical form
`
`triples express the semantic relationship of the form [object, property]. Ex. 1003
`
`¶221; Pet. 47.
`
`14
`
`
`
`
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`b. The claims do not require “determination between a class
`of noun.”
`To the extent Uniloc contends that the claims require “determination
`
`between a class of noun,” POR 21-21, Uniloc is wrong. Uniloc points to a single
`
`passage in the ’908 patent for the proposition that “the class of noun determines the
`
`object or the property of the keyfacts.” Id., 21. Uniloc fails to explain where the
`
`claims recite this requirement.
`
`Further, the passage does not require such a determination. Instead, the
`
`passage refers to a scenario where there is “a sequence of words having two or
`
`three nouns in a row,” and in that specific scenario, the ’908 patent explains that “it
`
`is likely” that certain nouns are objects while others are properties. ’908 patent,
`
`5:37-44.
`
`Demonstrating the flawed nature of Uniloc’s proposition, the ’908 patent
`
`provides an example in which the words “retrieval” and “information” are each
`
`tagged as “NV (vocative noun),” which according to Uniloc and the passage it
`
`cites would mean that both words are “properties.” POR 21. But, according to the
`
`’908 patent, one could be placed as an object and one as a property, even though
`
`they are both tagged as “vocative nouns”:
`
`
`
`15
`
`
`
`
`Excerpt of Table 1, ’908 patent, 6:19-22. Thus, Uniloc’s proposition has no
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`support in the specification.
`
`c. Braden-Harder determines between a class of noun.
`Assuming arguendo that the claims require “determination between a class
`
`of noun,” Braden-Harder renders this obvious. In the ’908 patent, the nouns may
`
`be further divided based on the class of noun and assigned a respective noun tag.
`
`’908 patent, 5:32-34. However, “the sequence of nouns is converted into a [single]
`
`keyfact tag KEY.” Id., 5:55-58; see also POR 2 (“into a distinct keyfact tag
`
`‘KEY’”). In other words, even if each noun is tagged with a different noun tag, the
`
`entire sequence of nouns (and their associated tags) is simply converted into a
`
`single tag.
`
`Contrary to Uniloc’s characterization, Braden-Harder does distinguish
`
`nouns. As shown in Fig. 22 below, Braden-Harder (as explained in Heidorn)
`
`distinguishes nouns by attaching different noun tags to the different classes of
`
`noun:
`
`16
`
`
`
`
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`
`
`Heidorn, Fig. 22 (annotated); Pet. 42; Ex. 1003 ¶199. As shown in the figure, the
`
`sequence of nouns “person whom I” has different noun tags (e.g., NOUN, PRON)
`
`associated with each noun. Thus, Braden-Harder attaches a different noun tag
`
`based on the type of noun, similar to the ’908 patent.
`
`Dr. Jansen further explains that “Braden-Harder similarly generates an
`
`intermediate-level tag sequence for an input sentence by replacing a sequence of
`
`lower-level-tagged noun phrases with a single tag ‘NP,’” Ex. 1003 ¶200; Pet. 42-
`
`43, which is also shown in the figure above. Thus, Braden-Harder similarly
`
`converts the sequence of nouns into a single tag like the ’908 patent.
`
`Finally, when there is a sequence of nouns, such as in Braden-Harder’s
`
`noun-modifiers example—“I like shark fin soup bowls”—Braden-Harder extracts
`
`logical form triples “to represent [the] possible internal structure of the noun
`
`17
`
`
`
`
`compounds” (“shark fin soup bowls”). Braden-Harder, 14:37-41; Ex. 1003 ¶¶216-
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`217; Pet. 45-46. Thus, Braden-Harder makes “determination between a class of
`
`noun.”
`
`d. Braden-Harder’s logical form triples represent “a paired
`relationship” between an object and a property.
`According to Uniloc, the form of [object, property] represents “a paired
`
`relationship between an ‘object’ that is the head and a ‘property’ that is the
`
`modifier.” POR 2 (quoting ’908 patent, 6:38-44), 7 (same). As explained supra,
`
`Braden-Harder’s logical form triples express the semantic relationship of the form
`
`[object, property]. Ex. 1003 ¶221; Pet. 47. Indeed, Dr. Jansen explained that “each
`
`of Braden-Harder’s logical form triples ‘contains two node words ... linked by a
`
`semantic relationship therebetween.’” See, e.g., Ex. 1003 ¶176; Pet. 34; Braden-
`
`Harder, 14:3-4. As such, each logical form triple is in a form that represents “a
`
`paired relationship” between a pair of node words. See Pet. 34; Ex. 1003 ¶176. In
`
`the case of noun modifier, Braden-Harder’s logical form triples represent “a paired
`
`relationship” between an object that is the noun and a property that is the modifier,
`
`and are thus in the paired form of [object, property]. See ’908 patent, 1:7-12.
`
`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. POR
`
`22. Uniloc cites no evidence to support its argument. Uniloc also has identified
`
`18
`
`
`
`
`nothing in the specification or the prosecution history that would limit “keyfacts”
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`to at most two elements (e.g., object and property) and nothing more. Cf. ’908
`
`patent, Tables 1 (“[KEY2 KEY1, NIL]”), 3 (“[KEY1 KEY2 KEY3]”). Nor has
`
`Uniloc provided any extrinsic evidence. Thus, Braden-Harder “may contain
`
`elements in addition to those explicitly mentioned in the claim” and still invalidate
`
`the claim. See In re Skvorecz, 580 F.3d 1262, 1265, 1267-68 (Fed. Cir. 2009).
`
`Further, as Dr. Jansen explained, while Braden-Harder’s logical form triples
`
`include the “[semantic] relations as an express part of the tuple: [object, Mods,
`
`modifier],” these logical form triples, nonetheless, “exhibit the semantic
`
`relationship of the form [object, modifier].” Ex. 1003 ¶221; Pet. 47. Uniloc does
`
`not dispute that the logical form triples for noun modifiers exhibit the semantic
`
`relationship between an object and its modifier, or property. Nor did Uniloc
`
`provide any evidence to support its argument.
`
`Moreover, Dr. Jansen further explained that it would have been obvious to
`
`generate other logic forms (Braden-Harder, 11:51-55) that do not include the
`
`relations to enable “‘structural paraphras[ing]’ ([id.,] 25:49-60).” Ex. 1003 ¶221;
`
`Pet. 47. Indeed, Braden-Harder teaches that “a logical form can take on any one of
`
`a number of different forms” and “can readily utilize any other form.” Braden-
`
`Harder, 11:46-55. 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
`
`19
`
`
`
`
`increase matches with ‘sufficiently similar semantic content’ (25:51-52) … [and]
`
`IPR2020-00755 Petition
`U.S. Patent No. 6,366,908
`
`to ensure that all relevant documents are retrieved from the document collection.”
`
`Ex. 1003 ¶221; Pet. 47. Uniloc does not dispute or address this.
`
`Braden-Harder renders obvious the claimed “keyfact,” even under Uniloc’s
`
`new construction.
`
`B. Uniloc’s rehashes of its POPR arguments do not call into question
`how Braden-Harder renders the claims obvious.
`1. Uniloc’s arguments that Braden-Harder is a phrase-based text
`retrieval method, POR 8, 22, fail.
`The Board previously considered and rejected Uniloc’s POPR arguments
`
`that Braden-Harder’s logic form retrieval method is a “phrase-based scheme.”
`
`Specifically, the Board rejected that “Braden-Harder’s logical form triples are the
`
`result of a phrase-based text retrieval method distinguishable from ‘extracting
`
`keyfacts of said document collection and said user