`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`APPLE INC.,
`Petitioner,
`v.
`PARUS HOLDINGS, INC.,
`Patent Owner.
`
`Case No. IPR2020-00686
`U.S. Patent No. 7,076,431
`
`PATENT OWNER’S SUR-REPLY TO
`PETITIONER’S REPLY TO PATENT OWNER’S RESPONSE
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`
`I.
`II.
`
`TABLE OF CONTENTS
`INTRODUCTION ........................................................................................... 1
`LADD DOES NOT TEACH OR DISCLOSE LIMITATION 1(C):
`“SPEAKER-INDEPENDENT SPEECH RECOGNITION DEVICE”........... 2
`A.
`Parus’s Construction of “Speaker-Independent Speech
`Recognition Device” is Consistent with the District Court’s
`Construction and the Intrinsic Record ................................................... 3
`1.
`Parus’s construction is consistent with the District
`Court’s construction. ................................................................... 4
`The plain and ordinary meaning should be afforded to the
`term “voice pattern” .................................................................... 4
`The Plain Disclosure of Ladd Demonstrates that Speech
`Recognition in Ladd is Dependent on Voice Patterns .......................... 6
`1.
`Apple’s identified speech recognition device operates in
`a manner proscribed by the ’431 Patent ...................................... 7
`Ladd’s use of a speech to text unit further undermines
`Apple’s argument ........................................................................ 8
`Apple’s Attempt to Conflate Speech Recognition with Natural
`Language Processing is Not Tenable .................................................. 10
`1.
`Speech Recognition does not include natural language
`processing .................................................................................. 10
`Mr. Occhiogrosso’s testimony does not support Apple’s
`two-step speech recognition theory .......................................... 13
`Ladd’s Disclosed “Voice Patterns” Are The Same “Voice
`Patterns” Excluded By The District Court’s Construction ................. 15
`III. MR. OCCHIOGROSSO’S OPINIONS SHOULD BE GIVEN
`GREATER WEIGHT THAN DR. TERVEEN’S ......................................... 18
`IV. CLAIM LIMITATION 1(K) ......................................................................... 19
`
`B.
`
`C.
`
`D.
`
`2.
`
`2.
`
`2.
`
`i
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`V. MOTIVATION TO COMBINE .................................................................... 21
`VI. CLAIMS 5-6 .................................................................................................. 23
`VII. CONCLUSION .............................................................................................. 23
`
`ii
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`
`TABLE OF AUTHORITIES
`
` Page(s)
`
`Cases
`ATD Corp. v. Lydall, Inc.,
`159 F.3d 534 (Fed. Cir. 1998) ............................................................................ 22
`In re Enhanced Security Research, LLC,
`739 F.3d 1347 (Fed. Cir. 2014) .......................................................................... 22
`In re Fine,
`837 F.2d 1071 (Fed. Cir. 1988) .......................................................................... 22
`In re Gorman,
`983 F.2d 982 (Fed. Cir. 1991) ............................................................................ 22
`
`iii
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`
`TABLE OF EXHIBITS
`
`Exhibit
`2001
`
`2002
`2003
`2004
`2005
`2006
`
`2007
`2008
`2009
`2010
`2011
`
`2012
`2013
`2014
`2015
`2016
`2017
`2018
`2019
`2020
`2021
`2022
`2023
`2024
`2025
`
`
`
`
`Description
`Order Denying Motion to Stay Pending Inter Partes Review, C.A. No.
`6:18-cv-00207-ADA
`Exhibit A3 Ladd Claim Chart 7076431
`Exhibit C Obviousness Claim Chart 7076431 (Corrected)
`Reserved
`Reserved
`Standing Order Regarding Scheduled Hearings in Civil Cases, 6:19-
`cv-00432-ADA
`Claim Construction Order, 1:20-cv-00351-ADA
`Claim Construction Order, 6:19-cv-00532-ADA
`Claim Construction Order, 6:18-cv-00308-ADA
`U.S. Patent No. 6,157,705 (Perrone)
`“instruction set” excerpt from 1997 Novell’s Dictionary of
`Networking
`Defendants’ Opening Claim Construction Brief, 6:19-cv-00432-ADA
`Excerpt of Case Docket Sheet, 6-19-cv-00278-ADA
`Excerpt of Case Docket Sheet, 6-19-cv-00514-ADA
`Excerpt of Case Docket Sheet, 6-19-cv-00515-ADA
`Excerpt of Case Docket Sheet, 7-18-cv-00147-ADA
`Markman Hearing Transcript, 6:19-CV-00432-ADA
`10/2/2020 Email to Court
`Excerpt of Case Docket Sheet, 6:18-cv-00308-ADA
`5/30/2019 Order Denying Stay, C.A. No. 6:18-cv-00207
`6/23/2020 Order Denying Stay, C.A. No. 6:19-cv-00514
`6/23/2020 Order Denying Stay, C.A. No. 6:19-cv-00515
`7/22/2020 Order Denying Stay, C.A. No. 7:18-cv-00147
`December 16, 2020 Deposition Transcript of Loren Terveen, Ph.D.
`Declaration of Benedict Occhiogrosso in Support of Patent Owner’s
`Response to Inter Partes Review of U.S. Patent No. 7,076,431
`
`
`
`
`
`
`iv
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`
`I.
`
`INTRODUCTION
`Apple’s primary argument for demonstrating that Ladd discloses the speaker-
`
`independent speech recognition device of claim 1 of the ’431 Patent rests on Apple’s
`
`new theory that speech recognition is a two-step process, which runs counter to the
`
`understanding of those skilled in the art. Apple conflates speech recognition, which
`
`converts speech to text in the ’431 Patent and matches speech to pre-determined
`
`voice patterns in Ladd, with language processing, which attempts to ascribe meaning
`
`to the converted text. This conflation appears to stem from the fact that Apple
`
`attempts to confuse speech recognition with a speech recognition device as required
`
`by the claims of the ’431 Patent, which does both speech recognition and language
`
`processing. Because this new theory of speech recognition rests on Apple’s
`
`transparent attempts to confuse speech recognition and speech recognition devices,
`
`all Grounds fail.
`
`As Parus will show, Apple’s attempts to support their arguments by
`
`mischaracterizing the POR and Mr. Occhiogrosso’s testimony only weakens
`
`Apple’s own arguments. Apple’s citations to Mr. Occhiogrosso’s testimony
`
`undercut their arguments and show the flaws in those arguments clearly. For these
`
`reasons alone, Mr. Occhiogrosso’s testimony should be given greater weight than
`
`Dr. Terveens’s testimony.
`
`1
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Apple’s other arguments ignore Parus and Mr. Occhiogrosso’s arguments
`
`from the POR. Parus will further demonstrate there is no motivation to combine the
`
`various references and Apple’s arguments regarding speech recognition run counter
`
`to the content of the dependent claims and cannot be true. Parus has shown and will
`
`further show that Apple relies on impermissible hindsight for motivation to combine
`
`Ladd with Kurosawa and Goedken. For these reasons, all Grounds fail.
`
`II.
`
`LADD DOES NOT TEACH OR DISCLOSE LIMITATION 1(C):
`“SPEAKER-INDEPENDENT SPEECH RECOGNITION DEVICE”
`Apple’s entire argument that Ladd discloses the speaker-independent speech
`
`recognition device is predicated upon two assertions, neither of which is accurate,
`
`and because of this Apple’s entire argument subsequently fails.
`
`Apple’s first assertion is that Parus is forced into some sort of limited claim
`
`construction in order for Parus to demonstrate that Ladd does not disclose claim
`
`limitation 1(c). (Paper 19, 2-3). A closer examination of Parus’s proposed claim
`
`construction reveals that it is not limited in any meaningful way and is encompassed
`
`by the District Court’s construction that Apple now purports to adopt.
`
`Apple’s second assertion is that the speaker-independent speech recognition
`
`device as taught by the ’431 and Ladd is a two-step process and because Ladd, as
`
`Apple interprets it, uses voice patterns in the second step instead of the first step,
`
`2
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Ladd does not disclose the voice patterns excluded in the ’431 Patent.1 The
`
`disclosure of Ladd directly contradicts this assertion. Also, a closer examination of
`
`Apple’s purported two-step speech recognition process demonstrates that Apple
`
`conflates speech recognition with language processing or understanding.
`
`Because Apple’s entire argument as to why Ladd discloses limitation 1(c) is
`
`predicated on these two false assertions, Apple’s entire argument fails.
`
`A.
`
`Parus’s Construction of “Speaker-Independent Speech
`Recognition Device” is Consistent with the District Court’s
`Construction and the Intrinsic Record
`Apple’s suggestion that Parus is forced to propose a construction that would
`
`exclude Ladd from teaching the “Speaker-Independent Speech Recognition Device”
`
`is unfounded because Ladd does not teach, or disclose, the “Speaker-Independent
`
`Speech Recognition Device” under the District Court’s or Parus’s construction.
`
`(Paper 19, 2-3).
`
`Recognizing that the Board is not bound by the District Court’s construction,
`
`and that the Board may independently decide the construction of the “Speaker-
`
`Independent Speech Recognition Device,” Parus advanced what it believes is the
`
`appropriate construction of the term. See Paper 15 at 21-24.
`
`1 Parus notes that this is the first time that Apple has advanced this two-step speech recognition theory.
`
`3
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Parus’s construction is consistent with the District Court’s
`construction.
`Contrary to Apple’s assertion, Parus’s proposed construction is entirely in
`
`1.
`
`agreement with the District Court’s construction. As can be seen in the table below
`
`by the bolded text, the District Court’s construction includes Parus’s entire
`
`construction:
`
`District Court’s Construction
`speech recognition device that
`recognizes spoken words without
`adapting to individual speakers or
`using predefined voice patterns
`
`Parus’s Construction
`speech recognition device that
`recognizes spoken words without using
`predefined voice patterns
`
`A proposed construction that is entirely encompassed by the District Court’s
`
`construction does not constitute a construction that Parus was forced to adopt as
`
`Apple contends. (Paper 19, 2-3). As will be demonstrated further below, Ladd does
`
`not teach, or disclose, the “Speaker-Independent Speech Recognition Device,” under
`
`either construction.
`
`2.
`
`The plain and ordinary meaning should be afforded to the
`term “voice pattern”
`Apple’s assertion that “Parus’s POR and supporting evidence provide no
`
`definition or explanation of what a ‘voice pattern’ is as used in Parus’s claim
`
`construction” is baseless. (Paper 19, 3). Parus believes that the plain and ordinary
`
`meaning of “voice pattern” should be afforded to the term, and that no further
`
`4
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`construction should be necessary. Parus’s use of the term in the POR and its
`
`supporting evidence support this belief.
`
`For example, in Mr. Occhiogrosso’s declaration in support of Parus’ POR,
`
`Mr. Occhiogrosso explains that Automatic Speech Recognition (“ASR”) “is a
`
`process by which a computer system recognizes the acoustic signal of human speech
`
`(converted by a microphone into an electrical signal) and translates this text of what
`
`was uttered.” (Ex. 2025, ¶ 49). Mr. Occhiogrosso further explains that ASRs can
`
`be broadly categorized as speaker-dependent and speaker-independent systems.
`
`(Ex. 2025, ¶ 50).
`
`Mr. Occhiogrosso then explains that speaker-independent systems can be
`
`further broken down into two categories. The first category is characterized by
`
`“pattern matching of incoming speech to a priori known templates of words is used
`
`in limited vocabulary” systems that seek to isolate specific commands. (Ex. 2025, ¶
`
`53). Mr. Occhiogrosso uses the Chigier reference to support this type of speaker-
`
`independent speech recognition. Id. Mr. Occhiogrosso notes that the second
`
`category is phoneme based that use statistical models “to extract [] Acoustic Features
`
`used by the acoustic model from the incoming speech after which statistical analysis
`
`estimates the likelihood that a particular sound (i.e. phoneme) is uttered.” Id. Mr.
`
`Occhiogrosso further explains that this category of speaker-independent speech
`
`recognition “allows for a much larger vocabulary and is very useful in creating a
`
`5
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`flexible system. Id. Mr. Occhiogrosso again relies on the Chigier reference to
`
`support this other type of speaker-independent speech recognition.
`
`Mr. Occhiogrosso then explains that the ’431 Patent disclaims the first
`
`category of speaker-independent speech recognition where the voice pattern is
`
`compared to a priori known templates of words used in a limited vocabulary, which
`
`results in a highly stilted ask and response type language, and favors a phoneme
`
`based speaker independent where larger vocabularies can be employed. (Ex. 2025,
`
`¶ 69). All of this is consistent with the plain and ordinary definition of the term
`
`“voice pattern.”
`
`B.
`
`The Plain Disclosure of Ladd Demonstrates that Speech
`Recognition in Ladd is Dependent on Voice Patterns
`Apple’s attempt to re-define Ladd’s speech recognition into a two-step
`
`process is a red herring and is undercut by the disclosure of Ladd. Speech
`
`recognition in Ladd is a single step process. The express disclosure of Ladd requires
`
`a speech recognition device that is based on voice patterns, something that is
`
`expressly disclaimed by the ’431 Patent. Therefore, Ladd’s speech recognition
`
`device cannot anticipate the speaker-independent speech recognition device of the
`
`’431 Patent.
`
`6
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Apple’s identified speech recognition device operates in a
`manner proscribed by the ’431 Patent
`A reading of Ladd shows that speech recognition is a single step process and
`
`1.
`
`that the difference between a speech recognition device based on voice patterns and
`
`the more flexible speech recognition device of the ’431 Patent is significant. The
`
`speech recognition device of Ladd performs speech recognition by recognizing voice
`
`patterns, not by converting speech to text based on phonemes. Following that
`
`recognition, the speech recognition device of Ladd matches those voice patterns to
`
`specific recognition grammars. It is the first step, performing speech recognition
`
`based on voice patterns or speech patterns that is proscribed by the ’431 Patent.
`
`Ladd does not convert speech to text, it recognizes voice patterns. Ladd
`
`describes its speech recognition device as follows: “[w]hen the ASR unit 254
`
`identifies a selected speech pattern of the speech inputs, the ASR unit 254 sends an
`
`output signal to implement the specific function associated with the recognized voice
`
`pattern.” (Ex. 1004, 9:36-39 (emphasis added)). The speaker–independent speech
`
`recognition device (of the ’431) performs speech recognition and language
`
`processing. There is no disclosure or teaching that the ASR unit 254 performs a first
`
`step of converting speech to text during speech recognition. The voice pattern is
`
`recognized and “an output signal to implement the specific function associated with
`
`the recognized voice pattern” is sent. Id. The speech recognition process of Ladd
`
`7
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`does not convert speech to text. Instead, it performs recognition based on voice
`
`patterns. The second step performed by the ASR unit 254 is identifying the signal
`
`to be sent based on the recognized voice pattern as identified by a recognition
`
`grammar. Id. Mr. Occhiogrosso’s deposition testimony is completely consistent
`
`with this understanding. (Ex. 1039, 43:16-44:7).
`
`The identified speech recognition device of Ladd is directly disclaimed by the
`
`’431 Patent. The ’431 Patent is crystal clear that the “speaker-independent speech
`
`recognition device” is not reliant on voice patterns. Specifically, the ’431 Patent
`
`states that: “The voice browsing system recognizes naturally spoken voice
`
`commands and is speaker-independent; it does not have to be trained to recognize
`
`the voice patterns of each individual user. Such speech recognition systems use
`
`phonemes to recognize spoken words and not predefined voice patterns.” (Ex. 1001,
`
`4:38-43; see also Ex. 1041, 2). The text of the references is clear and unambiguous.
`
`The ’431 Patent does not use “predefined voice patterns” and Ladd does.
`
`2.
`
`Ladd’s use of a speech to text unit further undermines
`Apple’s argument
`Apple’s argument, requiring that the speech recognition device first converts
`
`speech to text then recognizes it, is further undermined by Ladd’s disclosure of a
`
`speech to text unit 256 as part of the VRU server 234. Apple identifies the ASR unit
`
`254 of the VRU server 234 as satisfying the speaker-independent speech recognition
`
`8
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`unit of the claims of the ’431 Patent. (Paper 19 at 9-10). Apple also contends that
`
`“Ladd teaches the automatic speech recognition (ASR) unit 254 first recognizes the
`
`words from the user’s speech input and then performs the second step of determining
`
`whether the speech inputs match any key word or phrase via comparison to a stored
`
`grammar or vocabulary.” (Paper 19, 7). However, at no point does Ladd teach that
`
`the ASR unit 254 is capable of converting speech to text and Mr. Occhiogrosso
`
`provides no testimony that Ladd makes such a disclosure. See generally, Ex. 1039.
`
`The ASR unit 254 is described as being part of the VRU server 234 and having a
`
`particular operation, which does not include converting speech to text. (Ex. 1004,
`
`9:28-44). In the paragraph immediately following that description, the STT unit 256
`
`is described as being part of the VRU server 234 and having a particular operation,
`
`which does include converting speech to text. (Ex. 1004, 9:45-54). Notably, the
`
`STT unit 256 is not described as being speaker independent and then uses a different
`
`“preferable” software package called Dragon Naturally Speaking, which is
`
`commonly understood to be a speaker dependent system that requires extensive
`
`training. (
`
`). If, as now urged by Apple, the ASR unit 254 converted
`
`speech to text, there would be no reason to have a separate STT unit 256 that uses a
`
`different type of software to actually perform speech to text translation. The STT
`
`unit 256 is necessary because, as clearly disclosed by Ladd, the ASR unit 254 does
`
`not convert speech to text.
`
`9
`
`
`
`C.
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Apple’s Attempt to Conflate Speech Recognition with Natural
`Language Processing is Not Tenable
`In an attempt to escape the plain disclosure of Ladd, Apple argues that Ladd’s
`
`voice patterns are used at a different step in the speech recognition process, which
`
`runs counter to the plain disclosure of Ladd.2 Apple advances this new speech
`
`recognition theory in order to inaccurately claim that speech recognition requires
`
`two steps. (Paper 19, 5). According to Apple, step one comprises converting the
`
`spoken utterance into text or words, and step two determines the content of those
`
`words. Id. Apple then claims that Mr. Occhiogrosso agrees that speech recognition
`
`requires two steps. (Paper 19, 5-6). Speech recognition does not require two steps,
`
`and this is a mischaracterization of Mr. Occhiogrosso’s statements. (
`
`).
`
`1.
`
`Speech Recognition does not include natural language
`processing
`The speaker-independent speech recognition device of the ’431 Patent
`
`performs speech recognition and selection of the appropriate recognition grammar
`
`(i.e., language processing). Apple blurs the lines of speech recognition in an attempt
`
`to invalidate the claims of the ’431 Patent. Unfortunately for Apple, Ladd is clear
`
`2 To the extent the Board allows Dr. Terveen’s new opinions in his supplemental declaration, Parus submits Mr.
`Occhiogrosso’s supplemental declaration to rebut those new opinions.
`
`10
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`that its speech recognition device (the ASR unit 254) operates based on voice
`
`patterns, which is disclaimed by the ’431 Patent. (
`
`).
`
`The process that Apple describes as speech recognition includes speech
`
`recognition (step one of Apple’s process) and then natural language processing or
`
`understanding (step two of Apple’s process)—two distinctly different processes.
`
`Apple is attempting to confuse speaker-independent speech recognition, whether
`
`using voice patterns (Ladd) or phonemes (’431 Patent), with an automatic speech
`
`recognition device which includes both speech recognition and natural language
`
`processing (i.e., utilizing a set of grammars).
`
`As Mr. Occhiogrosso testified during his deposition, speech recognition as
`
`described by the ’431 Patent translates an audio input into text. (Ex. 1039, 51:19-
`
`25). This opinion is consistent with Mr. Occhiogrosso’s declaration where he
`
`testified that Automatic Speech Recognition (“ASR”) “is a process by which a
`
`computer system recognizes the acoustic signal of human speech (converted by a
`
`microphone into an electrical signal) and translates this into the text of what was
`
`uttered.” (Ex. 2025, ¶ 49). As described above, once the audio signal is translated
`
`into text, the speech recognition process is complete.
`
`According to the disclosure and claims of the ’431 Patent, after the speech
`
`recognition process is complete, natural language processing occurs to use the
`
`translated text to understand the speaker. This process may utilize grammars and is
`
`11
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`called language understanding or processing, which is entirely consistent with Mr.
`
`Occhiogrosso’s testimony. (Ex. 1039, 11:9-12:1, 64:20-66:3). In the natural
`
`language processing phase, the system of the ’431 Patent uses the translated text and
`
`attempts to match it with a recognition grammar that corresponds to potential
`
`capabilities of the system.
`
`The speech recognition device of Ladd, to the contrary, discusses recognizing
`
`a voice pattern and performing an action. (
`
`). The ASR unit 254 is
`
`described as “identif[ying] a selected speech pattern of the speech inputs” (speech
`
`recognition) and performing an action (activating a grammar). (Ex. 1004, 9:36-39;
`
`). Ladd provides additional description of speech recognition, which
`
`is also relied on by Apple. Ladd describes a “detection unit 260 records the audio
`
`inputs from the user and compares the audio inputs to the vocabulary of grammar
`
`stored in the database server unit 244.” (Ex. 1004, 10:13-15;
`
`).
`
`Ladd does not describe comparing the text derived from the audio input. Id. It
`
`describes comparing the “audio” to the grammar. (
`
`). Ladd matches
`
`predefined voice patterns to the audio input in the speech recognition phase, which
`
`is proscribed by the ’431 Patent. It does not convert audio to text and somehow
`
`compare that text to a voice pattern in a later phase.
`
`Apple’s interpretation of the speech recognition device of Ladd simply makes
`
`no sense. Ladd is unambiguous that “[w]hen the ASR unit 254 identifies a selected
`
`12
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`speech pattern of the speech inputs, the ASR unit 254 sends an output signal to
`
`implement the specific function associated with the recognized voice pattern.” (Ex.
`
`1004, 9:36-39;
`
`). If Apple’s interpretation is correct, Ladd would
`
`convert the audio input, which is a voice pattern, into text, then it would have to
`
`convert the text back to the voice pattern in order to match it to a particular voice
`
`pattern related to a grammar.
`
`Apple conflates these two different processes in an attempt to differentiate
`
`Ladd’s voice patterns from the voice patterns excluded from the ’431 Patent. (Paper
`
`19, 5-8). A POSITA would understand that Apple cannot claim that speech
`
`recognition requires two distinct steps because speech recognition is completed after
`
`step one of Apple’s purported two-step process. (
`
`).
`
`2. Mr. Occhiogrosso’s testimony does not support Apple’s
`two-step speech recognition theory
`In order to bolster their inaccurate two-step speech recognition theory, Apple
`
`claims that Mr. Occhiogrosso agrees with their theory. (Paper 19, 5-8). This is false
`
`and a mischaracterization of Mr. Occhiogrosso’s testimony. The testimony of Mr.
`
`Occhiogrosso identified by Apple highlights the difference between the speech
`
`recognition device of the ’431 Patent claims and the speech recognition unit (ASR
`
`254) of Ladd.
`
`13
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Almost all of the citations from Mr. Occhiogrosso’s deposition used by Apple
`
`concern testimony about the ’431 Patent, not Ladd. Mr. Occhiogrosso discusses the
`
`speech recognition device of the ’431 Patent as having two phases: conversion of
`
`speech to text, then applying the converted text to recognition grammars. (Ex. 1039,
`
`11:9-13, 15:4-20, 24:17-25:6, 29:14-23, 31:9-11, 32:24-33:1, 33:6-20, 38:5-39:2,
`
`40:2-22, 49:5-19, 50:17-52:22). None of this testimony applies to a speech
`
`recognition device in general or the speech recognition device of Ladd. It is all given
`
`as an explanation of what happens in the ’431 Patent.
`
`The citations to Mr. Occhiogrosso’s deposition related to a speech recognition
`
`device based on voice pattern matching provided by Apple demonstrate the
`
`difference between the ’431 Patent speech recognition device and the Ladd speech
`
`recognition device. Mr. Occhiogrosso states that the speech recognition device
`
`based on phonemes described by the ’431 Patent is fundamentally different than the
`
`speech recognition device based on voice patterns described by Ladd:
`
`Q. Do you have any opinion on whether there is
`a difference between recognizing a word via
`recognizing phonemes, a voice pattern, or artificial
`intelligence?
`A. Yes.
`
`Q. Are those three different methods of
`recognizing words?
`A. They're three different classes of
`algorithm.
`
`14
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`(Ex. 1039, 54:24-55:7). Mr. Occhiogrosso goes on to distinguish a speech
`
`recognition device based on voice pattern matching by staying that “[t]he voice
`
`pattern is simply captured in its entirety and then matched to a reference voice
`
`pattern.” (Ex. 1039, 43:16-44:7). It is uncontested that the speech recognition
`
`device of Ladd is reliant on voice pattern matching. See generally, Paper 19. As
`
`can clearly be seen, Mr. Occhiogrosso does not opine that a speech recognition
`
`device based on voice pattern matching must first convert speech to text.
`
`Finally, Mr. Occhiogrosso’s comments about the speech recognition device
`
`and the recognition grammars being uncorrelated is consistent with his opinion.
`
`Putting aside the fact that Mr. Occhiogrosso was not talking about Ladd specifically,
`
`speech recognition happens first and use of grammars happens second. Speech
`
`recognition and use of grammars is uncorrelated because the two processes happen
`
`in series. First, the speech recognition translates speech into text in the case of the
`
`’431 Patent or identifies a recognized voice pattern in the case of Ladd, then the
`
`appropriate recognition grammar is determined.
`
`D.
`
`Ladd’s Disclosed “Voice Patterns” Are The Same “Voice
`Patterns” Excluded By The District Court’s Construction
`Apple’s argument that Ladd’s disclosed “Voice Patterns” are somehow
`
`different than those contemplated by Parus’s and the District Court’s construction,
`
`which Apple has adopted, does not make logical sense for several reasons.
`
`15
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`First, Apple’s argument that because Ladd’s voice patterns are key words or
`
`key phrases makes them different than the excluded voice patterns of the ’431 Patent
`
`is not tenable. (Paper 19, 8). The ’431 Patent describes recognizing key words. (Ex.
`
`1001, 6:44-56). Ladd provides many examples of users using a single word to move
`
`onto the next step of the dialogue with Ladd’s IVR system. For example, at the
`
`initial state of the Ladd’s IVR, “[t]he user can respond with a choice of ‘weather’,
`
`‘market’, ‘news’ or ‘exit,’” in order to select the weather, market, or news
`
`application, or to just exit. (Ex. 1004 38:4-11). The system only recognizes the
`
`specifically allowed keywords. During his deposition, Dr. Terveen agreed that Ladd
`
`discloses applications that step through a prompt in which a user enters a selection
`
`at each step. (Ex. 2024, 54:23-56:18). The ’431 Patent recognizes these keywords
`
`based on phonemes, not voice patterns. Ladd recognizes the keywords based on
`
`voice patterns. The voice patterns of both references are the same.
`
`Second, Apple again argues that Ladd’s speech patterns cannot be the same
`
`as the excluded speech patterns from the ’431 Patent because Ladd detects the speech
`
`patterns using a grammar. (Paper 19, 10-11). Apple’s argument fails. Ladd
`
`discloses using grammars based on voice patterns. (Ex. 1004, 8:23-25, 9:32-39, and
`
`10:3-21). The ’431 Patent discloses using grammars based on phonemes, not voice
`
`patterns. (Ex. 1001, 4:30-58, 6:21-56).
`
`16
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Finally, Dr. Terveen’s arguments that Ladd’s voice patterns are different than
`
`the excluded voice patterns from the ’431 Patent are contradictory to Dr. Terveen’s
`
`and Apple’s arguments. For example, Dr. Terveen contends that Ladd’s speech
`
`patterns are key words or phrases determined using a grammar. (Ex. 1040, 18). But
`
`the examples cited by Dr. Terveen demonstrate that speech patterns would have to
`
`be analyzed at the speech recognition process step, not the natural language
`
`processing step that Dr. Terveen, and Apple, allude to when they discuss Ladd’s
`
`comparing speech commands to grammars.
`
`In one example, Dr. Terveen claims that Ladd was advanced enough to select
`
`a grammar based on the accent of the caller. (Ex. 1040, 17). But, a POSITA would
`
`understand that if Ladd is comparing a textual keyword to a grammar, as Dr. Terveen
`
`and Apple contend, then there would be no way to determine if the user has an
`
`accent. The only way to determine if the user had an accent would be at the speech
`
`recognition step when the voice pattern is analyzed, but Apple and Dr. Terveen
`
`contend that this does not occur in Ladd’s IVR system.
`
`Similarly, Dr. Terveen claims that Ladd can choose a speech recognition
`
`model based on the user’s speech patterns. (Ex. 1040, 16). Dr. Terveen indicates
`
`that based on the user’s speech patterns Ladd’s system could select the Spanish
`
`accent model, or the English accent model. Id. But if Ladd’s speech patterns are
`
`keywords as contended by Dr. Terveen and Apple, and the keyword is compared to
`
`17
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`a grammar after it has been converted to text in step one of Apple’s two step
`
`recognition theory, there would be no way for the system to detect an accent. For
`
`example, if Dr. Terveen, who has an American accent, spoke the word “boat” to
`
`Ladd’s IVR system, when Ladd compared the word “boat” to a grammar, there
`
`would be no way of determining if Dr. Terveen has an American or English accent.
`
`Similarly, if a British person spoke “boat” to Ladd’s IVR system, when Ladd
`
`compared the text word “boat” to a grammar, there would be no way of determining
`
`if the British person has an English or American accent based on comparing the text
`
`word “boat” with a grammar. Because Ladd’s voice patterns cannot be analyzed at
`
`the second step of Apple’s purported two-step theory of speech recognition, and
`
`because the Ladd system discloses stilted speech consisting of singular commands,
`
`Ladd’s voice patterns are the very voice patterns excluded from the ’431 Patent.
`
`III. MR. OCCHIOGROSSO’S OPINIONS SHOULD BE GIVEN
`GREATER WEIGHT THAN DR. TERVEEN’S
`Mr. Occhiogrosso’s opinions have been consistent
`
`throughout
`
`this
`
`proceeding. In contrast, Dr. Terveen opinions have shifted markedly. Dr. Terveen
`
`had ample opportunity to argue his and Apple’s new two-step speech recognition
`
`theory in his initial declaration. He chose to introduce this new theory only after he
`
`and Apple realized that Ladd’s speech recognition is the very type of speech
`
`recognition excluded by the ’431 Patent.
`
`18
`
`
`
`Case No. IPR2020-00686
`Patent No. 7,076,431
`Apple consistently mischaracterizes Mr. Occhiogrosso’s