`
`Google LLC,
`Petitioner,
`v.
`Uniloc 2017 LLC,
`Patent Owner.
`
`IPR2020-00755—Patent No. 6,366,908
`
`1
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`GOOGLE EXHIBIT 1052
`GOOGLE v. UNILOC
`IPR2020-00755
`
`
`
`Limited Issues: No Separate Arguments for 7/8 Instituted Grounds
`
`2
`
`Institution Decision 10
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Limited Issues: No Need to Decide
`☒ Level of Ordinary Skill
`☒ Step-Plus-Function Terms
`☒ Claim Construction of the Disputed Terms
`☒ Incorporation of the Heirdorn and Messerly patent applications
`
`3
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Limited Issues: Google’s Expert Testimony Remains Unrebutted
`
`• Uniloc has provided no expert declaration to rebut Dr. Jansen’s expert
`testimony
`• Uniloc has not sought to depose or cross-examine Dr. Jansen
`• Uniloc has provided no new evidence to change the Board’s reasonings in the
`Institution Decision
`
`4
`
`See generally Ex. 1003, Jansen Decl.
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Overview of Claim 6
`
`[6a]
`
`[6b]
`
`[6c]
`
`5
`
`Petition 17
`
`’908 patent, Fig. 1
`
`’908 patent, Fig. 2
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Constructions of the Term “Keyfact”
`
`Claim Term
`
`Google’s IPR
`Construction
`
`Uniloc’s Litigation
`Construction
`
`“keyfact” (all claims)
`
`Google proposed using
`Uniloc’s litigation
`construction in this IPR
`
`“Fact contained in
`sentences”
`
`Uniloc’s New
`Construction
`Raised in POR
`“a factual extraction of a
`sentence which
`expresses semantic
`relation between words
`in the sentence in the
`form of [object,
`property].”
`
`6
`
`Petition 23; Reply 2-3
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Uniloc’s New Construction vs. Braden-Harder
`• Uniloc proposes that “keyfact” means “a factual extraction of a sentence which expresses semantic relation
`between words in the sentence in the form of [object, property].” POR 9
`
`Petition 34
`
`Petition 47
`
`Petition 45
`
`7
`
`Petition 34, 45, 47; Reply 13-14
`
`’908 patent, 1:6-12
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`•
`
`Google’s Unrebutted Expert Testimony on “Keyfact”
`Expert, Dr. Jim Jansen
`“[A] person of ordinary skill in the art would have understood that Braden-Harder generates
`logical form triple for ‘the case of a constituent with noun modifiers’ (see Table 3), such as ‘Nadj’
`and ‘Mods’ relations that exhibit the semantic relationship of the form [object, modifier] and
`includes these relations as an express part of the tuple: [object, Mods, modifier].” Jansen Decl. ¶
`221.
`“Because Braden-Harder generates logical form triples for noun modifiers, Braden-Harder
`generates keyfacts having the forms of [object, modifier].” Jansen Decl. ¶ 221.
`
`•
`
`8
`
`Petition 45-47; Reply 10-13
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Google’s Unrebutted Expert Testimony on “Keyfact”
`
`Expert, Dr. Jim Jansen
`
`•
`
`“[I]t would have been obvious to generate other logic forms … that do not include the relations.”
`Jansen Decl. ¶ 221.
`• The modification is taught by Braden-Harder:
`
`Braden-Harder, 11:51-55
`
`Braden-Harder, 25:49-54
`
`9
`
`Petition 43, 47; Reply 10-13, 19-20
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Google’s Unrebutted Expert Testimony on “Keyfact”
`
`Expert, Dr. Jim Jansen
`
`• The modification would have worked together:
`– “In my opinion, a person of ordinary skill in the art would have been motivated to implement
`structural paraphrase for ‘noun appositive’ or ‘relative clause’ constructions to ensure that all
`relevant documents are retrieved from the document collection.” Jansen Decl. ¶ 221.
`– “A person of ordinary skill in the art would have been capable of implementing Braden-Harder
`to extract in other logic forms because doing so would require minimal modification without
`undue experimentation.” Jansen Decl. ¶ 221.
`– “Such a combination would have been no more than combining prior art elements according to
`known methods to yield predictable results.” Jansen Decl. ¶ 221.
`
`10
`
`Petition 47; Reply 10-13
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step
`
`’908 patent, 4:8-10
`
`11
`
`Petition 17-19
`
`’908 patent, Fig. 3
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step
`
`’908 patent, 5:15-19; see also 5:20-60
`
`’908 patent, 5:49-53
`
`’908 patent, 6:1-4; see also 5:61-6:4
`
`’908 patent, 6:34-37; see also 6:5-12, Table 1, 6:34-37
`
`12
`
`Petition 18-19; 25-27
`
`’908 patent, 6:45-51; see also 6:37-55
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Wherein the step of … comprises the steps of:
`Realtime Data, LLC
`
`’908 Patent
`
`Realtime Data, LLC, v. Iancu,
`Federal Circuit (Jan. 10, 2019)
`
`13
`
`Reply 5-8
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extraction Step: Analyzing a Document Collection
`Expert, Dr. Jim Jansen
`“Braden-Harder’s triple generation process 1100 separately ‘analyzes the
`textual phrases in the document and … constructs and stores a
`corresponding set of logical form triples, for that document, within dataset
`1030.’” Jansen Decl. ¶ 181
`
`14
`
`Petition 35-36
`
`Braden-Harder, Fig. 10A
`
`Braden-Harder, Fig. 11
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Analyzing a User Query
`Expert, Dr. Jim Jansen
`“Braden-Harder teaches a similar retrieval process (step 1245) for
`analyzing a user query to yield logical form triples.” Jansen Decl. ¶ 181
`
`15
`
`Petition 36-38
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`Braden-Harder, Fig. 12A
`
`
`
`[6a] Keyfact Extracting Step: Morphological Analysis
`Expert, Dr. Jim Jansen
`“Braden-Harder’s morphological analysis stems, or “normalize[s]
`differing word forms, e.g., verb tense and singular-plural noun
`variations, to a common morphological form for use by a parser”
`Jansen Decl. ¶ 185
`
`16
`
`Petition 3, 37-38, 39-40
`
`Braden-Harder, Table 1
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Extracting Keywords
`Expert, Dr. Jim Jansen
`“[I]n my opinion, Braden-Harder’s syntactic parse tree extracts words
`and phrases with no part-of-speech ambiguity.” Jansen Decl. ¶ 194
`“Heidorn incorporated by reference into Braden-Harder describes how
`syntactic processing extract keywords without part-of-speech ambiguity.”
`Jansen Decl. ¶ 195
`
`17
`
`Petition 39-43
`
`Heidorn, Fig. 22
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Part-of-Speech Tagging
`Expert, Dr. Jim Jansen
`“Braden-Harder’s parser then “syntactically analyze[s]”
`the stemmed words “using grammatical rules and
`attributes” (part-of-speech tags) to yield a syntactic
`parse tree.” Jansen Decl. ¶ 191
`
`18
`
`Petition 37-40
`
`Braden-Harder, Table 1
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Part-of-Speech Tagging
`Expert, Dr. Jim Jansen
`“By applying these syntax rules, Heidorn disambiguates
`any words having part-of-speech ambiguity and converts
`the initial set of part-of-speech tags into a final sequence
`of tags.” Jansen Decl. ¶ 198
`
`19
`
`Petition 37-38, 40-43
`
`Heidorn, Fig. 22
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Extracting Keyfacts
`Expert, Dr. Jim Jansen
`“Braden-Harder processes the syntactic parse tree
`using a set of semantic rules (‘keyfact pattern rules’) to
`yield a logical form graph (‘keyfact pattern’).” Jansen
`Decl. ¶ 205
`“Braden-Harder then uses at least four ‘graph walk’
`rules to generate a list of logical form triples. In my
`opinion, the logical form graph corresponds with the
`keyfact pattern and the graph walk rules correspond
`with the keyfact pattern generation rules. These logical
`form triples are extracted based on the semantic
`relationships of the nouns (i.e., keywords) with other
`words in the input string.” Jansen Decl. ¶ 211
`“In my opinion, these logical form triples represent
`semantic relationship between words and correspond
`to the claimed keyfacts.” Jansen Decl. ¶ 219
`
`•
`
`•
`
`•
`
`20
`
`Petition 43-47
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Keyfact Pattern Extraction
`Expert, Dr. Jim Jansen
`Braden-Harder incorporates Heidorn to “process[] the
`syntactic parse tree using a set of semantic rules (‘keyfact
`pattern rules’) to yield a logical form graph (‘keyfact
`pattern’).” Jansen Decl. ¶ 205
`
`21
`
`Petition 43-45
`
`Heidorn, Fig. 43
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6a] Keyfact Extracting Step: Keyfact Generation
`Expert, Dr. Jim Jansen
`“Braden-Harder then uses at least four ‘graph
`walk’ rules to generate a list of logical form
`triples” Jansen Decl. ¶ 211
`
`22
`
`Petition 45-47
`
`Braden-Harder, Fig. 5D
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6b] Keyfact Indexing Step
`
`23
`
`Petition 27-29
`
`’908 patent, Fig. 4
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6b] Keyfact Indexing Step
`
`’908 patent, 6:58-60
`
`’908 patent, 6:65-67
`
`24
`
`Petition 27-29
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`‘’908 patent, 7:1-9; see also 6:58-7:35
`
`
`
`Wherein the step of … comprises the steps of:
`Realtime Data, LLC
`
`’908 Patent
`
`Realtime Data, LLC, v. Iancu,
`Federal Circuit (Jan. 10, 2019)
`
`25
`
`Reply 5-8
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6b] Keyfact Indexing Step
`
`•
`
`•
`
`Expert, Dr. Jim Jansen
`“Braden-Harder teaches calculating
`‘predefined metrics for scoring documents
`… to provide enhanced document
`selectivity (discrimination). … [T]he
`scoring metrics may ‘take into account …
`the frequency of specific logical forms (or
`paraphrases thereof) and/or of particular
`logical form triples as a whole in that
`document.’” Jansen Decl. ¶ 229
`“Braden-Harder’s document indexing
`engine generates a Master Dataset 1030,
`which is an index structure.” Jansen Decl. ¶
`238
`
`Braden-Harder, Fig. 10A
`
`26
`
`Petition 48-54
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Claim 10
`
`•
`
`•
`
`•
`
`Expert, Dr. Jim Jansen
`“Messerly, which has been incorporated into Braden-
`Harder, teaches calculating the frequencies (tf) of
`various keyfacts in a document collection and
`document frequencies (by calculating idf) for the
`keyfacts.” Jansen Decl. ¶ 333
`“Grossman teaches generating a document index table,
`a document table, and a keyfact index table of said
`document collection as claimed and taught in the ’908
`patent.” Jansen Decl. ¶ 335
`“Grossman teaches generating INDEX, DOC, and
`TERM tables, which include among other things a
`document frequency (e.g., idf in the TERM table),
`document identifiers (e.g., DocId in the DOC table),
`and keyfact frequency (e.g., TermFrequency in the
`INDEX table).” Jansen Decl. ¶ 358
`
`27
`
`Petition 70-79
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[10] Tables for Index Structure
`Expert, Dr. Jim Jansen
`“[I]t would have been obvious to a person of ordinary skill in
`the art that in Grossman the DOC, INDEX, and TERM tables in
`Tables 5.5, 5.8, and 5.6 respectively correspond to the
`document table, document index table, and keyfact index table
`in the ’908 patent.” Jansen Decl. ¶ 337
`
`28
`
`Petition 71-78
`
`’908 patent, Fig. 4 (step 44)
`
`Grossman, 172,175
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6c] Keyfact Retrieving Step
`
`29
`
`Petition 29-31
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6c] Keyfact Indexing Step
`
`’908 patent, 7:35-40
`
`’908 patent, 7:41-45; see also 8:1-19
`
`30
`
`Petition 29-31
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`8:29-31
`
`
`
`Wherein the step of … comprises the steps of:
`Realtime Data, LLC
`
`’908 Patent
`
`Realtime Data, LLC, v. Iancu,
`Federal Circuit (Jan. 10, 2019)
`
`31
`
`Reply 5-8
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[6c] Keyfact Retrieving Step
`
`Expert, Dr. Jim Jansen
`“Braden-Harder … generates logical form triples from the user
`query” and it “teaches receiving the logical form triples
`associated with the retrieved document records.” Jansen Decl. ¶
`256.
`“Braden-Harder’s weights are keyfact weight constants in
`accordance with said keyfact pattern. As such, it is my opinion
`that Braden-Harder defines a retrieval model in consideration of
`weight constants according to logical form triples, like the ’908
`patent.” Jansen Decl. ¶ 265
`“Braden-Harder teaches displaying the retrieved documents in
`rank order according to similarity.” Jansen Decl. ¶ 278
`
`•
`
`•
`
`•
`
`32
`
`Petition 55-62, 84
`
`Braden-Harder, Fig. 12A
`
`Braden-Harder, Fig. 8B
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Claim 11
`
`•
`
`•
`
`•
`
`•
`
`Expert, Dr. Jim Jansen
`“Braden-Harder, when implemented with Grossman,
`uses a vector-space retrieval strategy that, as
`Grossman teaches, forms a vector Q for the terms
`found in the query and a vector D of size t (t being
`the number of terms in the document collection) for
`each document.” Jansen Decl. ¶ 366
`In Braden-Harder, “‘[t]he weight reflects a relative
`importance ascribed to that relation in indicating a
`correct semantic match between a query and a
`document’ and is ‘generally defined on an empirical
`basis.’” Jansen Decl. ¶ 369
`“[A] person of ordinary skill in the art would have
`recognized that Braden-Harder’s weight constants
`would be augmented with Messerly’s tf-idf weights,
`which would be stored in Grossman’s document and
`query vectors.” Jansen Decl. ¶ 373
`“Braden-Harder teaches displaying a retrieval result
`by applying said weights to its NLP-based IR
`model.” Jansen Decl. ¶ 378
`
`33
`
`Petition 79-83
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[11] Calculating Keyfact Weights
`
`Grossman, 16
`
`***
`
`34
`
`Petition 12, 48-49, 57-60
`
`Messerly, (1). (3)
`
`Braden-Harder, 17:60-18:7
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`[11] Keyfact Weight Constants
`
`35
`
`Petition 31, 58, 80-81; Reply 27-30
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`’908 patent, 7:41-64
`
`Braden-Harder 16:19-31
`
`
`
`[11] Keyfact Weight Constants
`Expert, Dr. Jim Jansen
`• “Braden-Harder’s weights are based on the different semantic relationships in the
`logical form triples, such as the words’ functional roles (Table 2), semantic roles (Table
`3), or semantic labels (Table 4).” Jansen Decl. ¶ 263
`• “[A]person of ordinary skill in the art would either multiply Braden-Harder’s static
`numeric weights with Messerly’s tf-idf weights or sum the two weights. By using both
`weights in combination, the person of ordinary skill in the art would take into
`consideration how often a logical form triple appears in relevant documents, the number
`of times it occurs in the whole collection, and its empirical weight compared to other
`triples in the collection. In my opinion, such a combination would have been no more
`than combining prior art elements according to known methods to yield predictable
`results.” Jansen Decl. ¶ 374
`
`36
`
`Petition 58-59, 82
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`“Disparaged Art”
`
`•
`
`“Uniloc observed that the Petition takes an overly
`expansive view of ‘keyfact’ that ‘would
`impermissibly compass disparaged art’ cited
`during prosecution. … Tellingly, Google does not
`deny these observations concerning either the
`construction applied in the Petition or its
`inconsistency with the prosecution history.” Sur-
`Reply 2
`
`37
`
`Sur-Reply 2; Petition 1, 91
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`Petition 1
`
`Petition 91
`
`
`
`Conceptual Based Schemes
`
`’908 patent, 1:45-49
`
`’908 patent, 1:55-59
`
`Expert, Dr. Jim Jansen
`• “During the third semantic phase of NLP, the text is
`checked for meaningfulness, and the various relationships
`between words (lexical items) are analyzed. Often the
`output of this phase is a context-free representation of the
`input sentence that translates the input sentence to logic
`(e.g., logical form).” Jansen Decl. ¶ 58
`• “Braden-Harder’s logical form triples provide a concept-
`based retrieval like the ’908 patent’s keyfacts.” Jansen Decl.
`¶ 176
`
`Braden-Harder discussing the limit of prior approaches:
`
`38
`
`Petition 17; 34
`Reply 21
`
`Braden-Harder, 4:18-24
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Keyfacts: Determination between classes of noun
`Expert, Dr. Jim Jansen
`• “[F]or the input string, ‘I like shark fin soup bowls,’ a list of
`eight logical form triples is generated.” Jansen Decl. ¶ 217
`• “In my opinion, these logical form triples represent
`semantic relationship between words and correspond to the
`claimed keyfacts.” Jansen Decl. ¶ 219
`
`Braden-Harder attaches a different noun tag
`based on the type of noun, similar to the ’908 patent:
`
`Heidorn, Fig. 22
`
`39
`
`Petition 45, 47; Reply 17-18
`
`Braden-Harder, 14:37-49
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Contradicts the ’908 patent’s definition of keyfact
`
`’908 patent, 1:15-16
`
`Uniloc’s New Construction
`
`An important factual extraction of a sentence
`which expresses semantic relation between words contained in the sentences
`in the form of [object, property].
`
`40
`
`Reply 3-4
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Uniloc argued that keyfact does not require any form
`
`Uniloc Opening Claim Construction Brief, 6
`
`41
`
`Reply 3-4
`
`Uniloc Reply Claim Construction Brief, 6
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Sur-Reply 2
`
`’908 patent, Table 1
`
`42
`
`Sur-Reply 2
`
`Sur-Reply 13
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Sur-Reply 13
`
`43
`
`Petition 27-28; Sur-Reply 1
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`’908 patent, Table 1
`
`’908 patent, Table 2
`
`
`
`Sur-Reply 13
`
`44
`
`Sur-Reply 13; Reply 18-19.
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`’908 patent, Table 2
`
`’908 patent, Table 3
`
`
`
`Motivation to Modify
`
`Messerly uses logical tokens with characters
`
`Braden-Harder 23:67-24:6
`
`45
`
`Sur-Reply 14-16, Petition 11
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Keyfact Weight Constants
`
`Expert, Dr. Jim Jansen
`
`Ex. 1003, Jansen Decl. ¶ 374
`
`46
`
`Petition 31, 58, 82; Reply 27-30
`
`’908 patent, 7:41-64
`
`Braden-Harder 16:19-31
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Keyfact Weight Constants
`
`47
`
`Petition 21; Sur-Reply 20
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`’908 patent, 7:41-64
`
`
`
`Keyfact Weight Constants
`
`Braden-Harder 25:41-48
`
`48
`
`Reply 29
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Keyfact Type
`
`49
`
`Petition 81; Reply 19
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`’908 patent, 8:10
`
`
`
`Sur-Reply 8
`
`Reply 5
`
`50
`
`Sur-Reply 8
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`POR 15
`
`POR 16
`
`
`
`Phrase-Based Scheme
`
`POPR 26
`
`POR 22
`
`51
`
`POPR 26, POR 22, Institution Decision 30-31
`
`Institution Decision 30-31
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`’908 Patent, Table 1
`
`52
`
`Petition 27
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Claim 7: Selecting a tag sequence out of sequences
`Expert, Dr. Jim Jansen
`• “Kucera teaches using morphological analysis to resolve
`ambiguities due to meaning and parts of speech. For
`example, Kucera teaches that when a sentence may be
`‘broken down into one or more strings of ambiguously-
`tagged words,’ the words may have ‘multiple tags.’” Jansen
`Decl. ¶ 395
`• “Kucera teaches using a ‘a collocational tag disambiguation
`processor [to] appl[y] an empirically-compiled probability-
`like function defined on adjacent pairs of syntactic tags to
`determine a unique sequence of tags (one for each word)
`corresponding to the most probable parse of each
`ambiguously-annotated word in the sentence.’ … Thus, out
`of ‘the ν possible tag sequences, a single sequence Xc is
`selected as most probably correct.’ … In my opinion, the
`selection of the most probable tag sequence in Kucera
`corresponds to this “selecting” step in claim 7.” Jansen
`Decl. ¶ 396
`
`53
`
`Petition 85-87
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Claim 8
`
`•
`
`•
`
`•
`
`•
`
`•
`
`Expert, Dr. Jim Jansen
`“Braden-Harder teaches parsing the input sentence into
`its constituent words using a stored lexicon.” Jansen
`Decl. ¶ 303
`“Braden-Harder’s morphological analysis … us[es] a
`stored lexicon.” Jansen Decl. ¶ 307
`“[I]nopinion, Braden-Harder teaches that its lexicon
`(machine-readable my dictionary) comprises multiple
`part-of-speech classes.” Jansen Decl. ¶ 309
`“Braden-Harder’s morphological analysis stems the
`words ‘to normalize differing word forms, e.g., verb
`tense and singular-plural noun variations, to a common
`morphological form for use by a parser.’” Jansen Decl.
`¶ 314
`“Braden-Harder’s NLP-based IR method obtains an
`initial set of part-of-speech tag sequence for the input
`text.” Jansen Decl. ¶ 318
`
`54
`
`Petition 66-68
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Claim 9
`
`Expert, Dr. Jim Jansen
`• “Braden-Harder teaches that its lexicon (machine-readable
`dictionary) comprises various classes of words — ‘prepositions,
`conjunctions, verbs, nouns, operators and quantifiers.’ Ex. 1020,
`Braden-Harder, 12:12-16, 14:20-24, Table 3 (including “adjectives”
`as one of the classes).” Jansen Decl. ¶ 322
`• “Heidorn, which Braden-Harder incorporates, explains that a
`dictionary may include part-of-speech information such as
`adjectives and adverbs.” Jansen Decl. ¶ 323
`• “Although [Braden-Harder] does not expressly use ‘a stop-word
`lexicon,’ Braden-Harder skips certain terms such as ‘a, an, the’
`when creating its logical form graphs. … It is my opinion that it
`would have been obvious to use Grossman’s list of stop-words
`because ‘[r]emoving these terms reduces index construction, time
`and storage cost….’ Jansen Decl. ¶ 325
`• “[T]here is no functional difference with using a single dictionary
`with multiple part-of-speech classes versus using multiple smaller
`dictionaries, each with its own part-of-speech class. … But to the
`extent Patent Owner disagrees, it would have been obvious to
`implement Braden-Harder’s paraphrasing with Miller’s WordNet
`synsets (synonym sets), which are stored in separate part-of-speech
`subnets (e.g., dictionaries).” Jansen Decl. ¶¶ 415-417
`
`55
`
`Petition 69, 90
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`
`
`Claim 12
`
`Expert, Dr. Jim Jansen
`“As Fig. 8B shows, Braden-Harder’s retrieval result
`indicates documents with at least one matching triple
`(18:8-20) and thus indicates documents having
`similar matching triples with the query triples.
`Therefore, it is my opinion that Braden-Harder
`renders ‘said retrieval result indicates documents
`with a keyfact similar to said keyfact of said user
`query’ obvious.” Jansen Decl. ¶ 383
`
`56
`
`Petition 84
`
`Braden-Harder, Fig. 8B
`
`DEMONSTRATIVE EXHIBIT—NOT EVIDENCE
`
`