`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
` UNITED STATES PATENT AND TRADEMARK OFFICE
`
`
`
`
`
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`
`
`
`
`
`APPLE INC.,
`Petitioner,
`
`v.
`
`Zentian Limited
`Patent Owner.
`____________________
`
`Case IPR2023-00037
`Patent No. 10,971,140
`____________________
`
`
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`
`
`
`
`
`
`
`
`
`IPR2023-00037
`PATENT OWNER’S PRELIMINARY RESPONSE
`TABLE OF CONTENTS
`
`
`I. Introduction
`II. Background of the ’140 Patent
`III. Person of ordinary skill in the art
`IV. Claim construction
`V. The Board should deny institution under Fintiv
`
`1
`2
`2
`2
`3
`
`4
`A. Fintiv factors 1-5 favor denial of institution
`B. Fintiv factor 6 requires a “compelling” challenge, which the Petition does
`6
`
`not present
`VI. An ordinary artisan would not have been motivated to substitute Chen’s
`supercomputer processing architecture in place of Jiang’s conventional processor
`for a personal computer
`VII. The Petition fails to demonstrate obviousness as to a shared acoustic model
`14
`memory containing acoustic model data for a cluster of processors
`VIII. The Petition fails to demonstrate obviousness as to “each of the plurality of
`processors [being] configured to compute a probability,” as recited in 1(c)
`18
`IX. Conclusion
`21
`
`
`6
`
`
`
`
`
`
`
`
`
`
`
`- 1 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`
`
`EXHIBIT LIST
`
`
`Description
`
`Intentionally Left Blank
`Declaration of Deliang Wang, Ph.D in support of Patent
`Owner’s Preliminary Response
`Zentian Ltd. v. Apple. Inc., No. 6:22-cv-00122-ADA, Apple’s
`Preliminary Invalidity Contentions
`Zentian Ltd. v. Apple. Inc., No. 6:22-cv-00122-ADA,
`Scheduling Order
`
`
`
`
`Exhibit No.
`2001
`
`2002
`
`2003
`
`2004
`
`
`
`
`
`
`
`- 2 -
`
`
`
`
`
`
`IPR2023-00037
`PATENT OWNER’S PRELIMINARY RESPONSE
`
`I.
`
`Introduction
`To meet the claims of the ’140 patent, the Petition argues that it would have
`
`been obvious to substitute the clustered processor and memory architecture from
`
`one specific embodiment in Chen—which requires at least eight processors in four
`
`clusters with four memories—in place of Jiang’s processor.
`
`But the Petition omits to address that Jiang’s “exemplary system” in its
`
`“Preferred Embodiment” is a “general purpose computing device in the form of a
`
`conventional personal computer 20,” with a conventional CPU, Ex. 1004, 5:4-7,
`
`whereas Chen’s highly complex eight-processor four-memory clustered design is
`
`for a supercomputer. Ex. 1005, 4:4-5, 4:57-5:6, 1:29-31.
`
`As this Preliminary Response demonstrates, there was nothing “obvious”
`
`about replacing Jiang’s conventional personal computing hardware with Chen’s
`
`supercomputer processing and memory architecture, and an ordinary artisan would
`
`not have been motivated to make such a fanciful combination.
`
`Nor would an ordinary artisan have been motivated to implement Jiang’s
`
`speech recognition teachings using Chen’s supercomputer processors and
`
`memories. An ordinary artisan would have understood that Jiang’s CPU-based
`
`speech recognition system was not suited for combination with Chen’s parallel
`
`processing system.
`
`
`
`
`
`
`
`
`
`- 1 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`Moreover, beyond simply combining Jiang’s teachings with Chen’s
`
`hardware, the Petition relies on further modifications to the combination of Jiang
`
`and Chen in order to meet specific requirements of the challenged claims. Those
`
`modifications, however, are not supported by any motivation; instead, they are
`
`simply hindsight-driven reconstructions of the challenged claims.
`
`The Petition thus fails to demonstrate a reasonable likelihood of success, and
`
`falls well short of the much more stringent “high likelihood” of success that the
`
`Board should apply here in view of Fintiv and recent precedents.
`
`II. Background of the ’140 Patent
`U.S. Patent 10,971,140, titled “Speech recognition using parallel
`
`processors,” is directed to an improved speech recognition circuit that “uses
`
`parallel processors for processing the input speech data in parallel.” Ex. 1001,
`
`1:18-20. The ’140 patent teaches multiple processors “arranged in groups or
`
`clusters,” with each group or cluster of processors connected to one of several
`
`“partial lexical memories” that “contains part of the lexical data.” Ex. 1001, 3:13-
`
`18. “Each lexical tree processor is operative to process the speech parameters using
`
`a partial lexical memory and the controller controls each lexical tree processor to
`
`process a lexical tree corresponding to partial lexical data in a corresponding
`
`partial lexical memory.” Ex. 1001, 3:19-24. The ’140 patent further teaches that
`
`the invention “provides a circuit in which speech recognition processing is
`
`
`
`
`
`
`
`
`- 2 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`performed in parallel by groups of processors operating in parallel in which each
`
`group accesses a common memory of lexical data.” Ex. 1001, 3:62-66.
`
`III. Person of ordinary skill in the art
`For purposes of this Preliminary Response, Patent Owner does not dispute
`
`the Petition’s proposed level of skill for a person of skill in the art (“POSA”),
`
`because the level of skill in the art is not necessary for addressing any disputes
`
`between the parties.
`
`IV. Claim construction
`The Petition does not propose any particular claim constructions and raises
`
`no claim construction issues.
`
`Patent Owner’s arguments in this Preliminary Response are based on the
`
`plain and ordinary meaning of the words of the claims themselves in view of the
`
`associated teachings of the specification. However, to the extent the Board views
`
`Patent Owner’s arguments as resting on claim construction positions, the bases for
`
`such positions are set forth in the corresponding argument sections below.
`
`V. The Board should deny institution under Fintiv
`Whereas inter partes review was created “to establish a more efficient and
`
`streamlined patent system that will improve patent quality and limit unnecessary
`
`and counterproductive litigation costs,” “[p]arallel district court and AIA
`
`proceedings involving the same parties and invalidity challenges can increase,
`
`
`
`
`
`
`
`
`- 3 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`rather than limit, litigation costs.” USPTO Memorandum, Interim Procedure for
`
`Discretionary Denials in AIA Post-Grant Proceedings with Parallel District Court
`
`Litigation (“Guidance Memo”) (June 21, 2022), at 1.
`
`In considering whether to institute an inter partes review proceeding where
`
`there is a parallel district court litigation, the Board considers the following six
`
`factors set forth in Apple Inc. v. Fintiv, Inc., IPR2020-00019, Paper 11 (PTAB
`
`Mar. 20, 2020) (precedential):
`
`1. Whether the court granted a stay or evidence exists that one may be
`
`granted if a proceeding is instituted;
`
`2. Proximity of the court’s trial date to the Board’s projected statutory
`
`deadline for a final written decision;
`
`3. Investment in the parallel proceeding by the court and the parties;
`
`4. Overlap between issues raised in the petition and in the parallel
`
`proceeding;
`
`5. Whether the petitioner and the defendant in the parallel proceeding are
`
`the same party; and
`
`6. Other circumstances that impact the Board’s exercise of discretion,
`
`including the merits.
`
`
`
`
`
`
`
`
`
`- 4 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`Fintiv factors 1-5 favor denial of institution
`A.
`Factor 1: The patent at issue in this IPR is the subject of pending litigation in
`
`Zentian Ltd. v. Apple, Inc., C.A. No. 6:22-cv-00122-ADA (W.D. Tex.). No stay
`
`has been granted, and neither party has sought a stay to the extent this proceeding
`
`is instituted.
`
`Factors 3 and 5: The parties in the district court case are the same as the
`
`parties in this proceeding. That case has continued to proceed since the Petition in
`
`this proceeding was filed, and will continue to proceed. The parties have
`
`exchanged proposed claim constructions, have completed claim construction
`
`briefing, and are awaiting a Markman hearing. Preliminary infringement and
`
`invalidity contentions have also been exchanged. Fact discovery will commence on
`
`March 24, 2023, and the parties are completing discovery on Apple’s pending
`
`motion to transfer to the Northern District of California.
`
`Factor 4: Apple’s preliminary invalidity contentions in the district court case
`
`rely on Jiang, Chen, Lucke, Robinson, and Wrench, which are all of the references
`
`at issue in this IPR. Compare Pet. 6 with Ex. 2003 at 26, 27-28.
`
`Factor 2: A trial date has tentatively been set for April 22, 2024, and will
`
`more definitively be set after the Markman hearing. Ex. 2004 at 5. By contrast, the
`
`Board’s Institution Decision in this proceeding is not due until June 15, 2023,
`
`placing the Board’s projected final written decision date at June 15, 2024—several
`
`
`
`
`
`
`
`
`- 5 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`months after the tentative trial date in the district court proceeding. Accordingly,
`
`there is a high likelihood that a final written decision in this inter partes review
`
`would come later than a decision in the district court litigation, and that the
`
`Board’s decision would merely address the same prior art and issues raised in the
`
`district court.
`
`Accordingly, Fintiv factors 1-5 favor discretionary denial.
`
`B.
`
`Fintiv factor 6 requires a “compelling” challenge, which the
`Petition does not present
`Because Fintiv factors 1-5 favor discretionary denial, the Board must
`
`consider whether the Petition presents “compelling merits.” Commscope Techs.
`
`LLC v. Dali Wireless, Inc., IPR2022-01242, Paper 23 at 4 (Feb. 27, 2023)
`
`(precedential).
`
`“The compelling merits standard is a higher standard than the standard for
`
`institution set by statute.” Id. at 3. To meet the “compelling merits” standard, the
`
`Petition must demonstrate that it is “highly likely that the petitioner would prevail
`
`with respect to at least one challenged claim.” Id. at 3-4 (emphasis added).
`
`As this Preliminary Response demonstrates, the Petition fails to demonstrate
`
`even a reasonable likelihood of success, much less the “high likelihood” of success
`
`standard that the Board should apply in view of Fintiv and Commscope.
`
`
`
`
`
`
`
`
`
`- 6 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`VI. An ordinary artisan would not have been motivated to substitute Chen’s
`supercomputer processing architecture in place of Jiang’s conventional
`processor for a personal computer
`Claim 1 of the ’140 patent is the sole challenged independent claim in this
`
`proceeding. Limitations 1(a) recites “one or more clusters of processors, each of
`
`the one or more clusters of processors comprising: a plurality of processors.” Pet.
`
`66. Jiang does not teach one or more clusters of processors, as the Petition admits.
`
`Pet. 14. To fill that gap, the Petition instead looks to Chen. Pet. 15. The Petition
`
`proposes to “substitute[]” Chen’s processing architecture, shown below, in place of
`
`Jiang’s disclosed processor. Pet. 19; Ex. 1004, Fig. 4 (annotated).
`
`
`
`
`
`
`
`
`
`- 7 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`
`
`
`The Petition’s combination, however, fails to address or even consider the
`
`highly disparate contexts of Chen and Jiang’s systems, and the fundamental
`
`incompatibility of Chen’s processing architecture as a substitute for Jiang’s
`
`processor. Ex. 2002 ¶ 25.
`
`
`
`
`
`
`
`
`
`- 8 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`Jiang teaches that, in its “Preferred Embodiment,” “an exemplary system for
`
`implementing the invention includes a general purpose computing device in the
`
`form of a conventional personal computer 20, including processing unit 21. . . .”
`
`Ex. 1004, 5:4-7; Ex. 2002 ¶ 26. Jiang likewise repeatedly refers to “personal
`
`computer 20” throughout its disclosure as the “exemplary system” in which its
`
`speech recognition teachings are to be implemented and usable. Ex. 1004, 5:13-16,
`
`5:27-30, 5:44-46, 5:55-58, 6:6-10, 6:12-15, 6:28-31, 6:39-42; Ex. 2002 ¶ 26.
`
`Indeed, Jiang specifically teaches: “tree search engine 74 is preferably
`
`implemented in CPU 21,” which is the CPU shown for personal computer 20, “or
`
`may be performed by a dedicated speech recognition processor employed by
`
`personal computer 20.” Ex. 1004, 6:39-42; Ex. 2002 ¶ 26. An ordinary artisan
`
`would have known at the time that Jiang’s CPU 21 for a conventional personal
`
`computer would have been of the nature of the Intel and AMD personal computer
`
`processors that dominated the market through the late 1990s and early 2000s. Ex.
`
`2002 ¶ 26. An ordinary artisan would have likewise understood that Jiang’s
`
`reference to “a dedicated speech recognition processor” referred to another
`
`processor of the same nature as the Intel or AMD processor that would have served
`
`as CPU 21, or else a digital signal processor of the type sold by Texas Instruments
`
`at the time. Ex. 2002 ¶ 26. And While Jiang teaches that CPU 21 in personal
`
`
`
`
`
`
`
`
`
`- 9 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`computer 20 “may include one or more processors,” Ex. 1004, 6:39-41, an
`
`ordinary artisan would have understood that teaching to refer to the possibility of
`
`using multiple hardware processors included in a conventional personal computer
`
`at the time to form CPU 21. Ex. 2002 ¶ 26.
`
`Chen, by contrast, is a patent disclosed by Cray Research, which was known
`
`at the time as a leading designer of cutting edge, extremely expensive and complex
`
`supercomputing systems. Ex. 1005 at 1 (“Assignee”); Ex. 2002 ¶ 27. Indeed, Chen
`
`expressly refers to the “original supercomputer developed by the assignee of the
`
`present invention.” Ex. 1005, 4:4-5; Ex. 2002 ¶ 17. Moreover, the Chen reference
`
`is expressly directed towards improving upon the prior supercomputers developed
`
`by Cray and others. Ex. 1004, 4:57-5:6; Ex. 2002 ¶ 27. Indeed, Chen expressly
`
`states “[t]he present invention relates generally to parallel processing computer
`
`systems for performing multiple-instruction-multiple-data (MIMD) parallel
`
`processing.” Ex. 1004, 1:29-31. A person of ordinary skill would have understood
`
`that “parallel processing computer systems for performing . . . MIMD parallel
`
`processing” at the time were supercomputers, not conventional personal
`
`computers. Ex. 2002 ¶ 27. Moreover, while Chen teaches that its disclosed parallel
`
`processing computer system could be built using “commercially [sic] single chip
`
`microprocessors, as well as commercially available memory chips,” Ex. 1005,
`
`
`
`
`
`
`
`
`
`- 10 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`10:18-21, Chen’s system would have nonetheless required four clusters of
`
`processors, each cluster containing at least two processors, and each cluster
`
`connected to a dedicated cluster shared memory, with multiple clusters also
`
`adjacently interconnected to another cluster’s memory. Ex. 1005, 9:10-19; Ex.
`
`2002 ¶ 27. In other words, Chen’s system required at least eight processors and
`
`four cluster shared memories, as well as complex intra-connections and
`
`interconnections between the processors themselves, the processors to various
`
`memories, and the memories to one another. Ex. 2002 ¶ 27. An ordinary artisan
`
`would have understood that, even using commercially available processors and
`
`memory chips, Chen’s architecture would have entailed far greater cost,
`
`complexity, and hardware space requirements than what would have been suitable
`
`for a conventional personal computer at the time. Ex. 2002 ¶ 27.
`
`Indeed, in the late 1990s and early 2000s, the costs of processors and
`
`memories were generally known to be the most significant cost categories for
`
`personal computers. Ex. 2002 ¶ 28. Thus, for instance, an ordinary artisan would
`
`have understood that adding seven more processors and three more memories and
`
`creating the necessary interconnections to form Chen’s supercomputer architecture
`
`in the context of a personal computing device would have rendered the resulting
`
`
`
`
`
`
`
`
`
`- 11 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`computing system cost prohibitive for the general public, and thus for the vast
`
`majority of automatic speech recognition customers. Ex. 2002 ¶ 28.
`
`Moreover, successfully modifying a standard personal computer to
`
`implement Chen’s supercomputer processor and memory architecture would have
`
`required a level of expertise in parallel computer design that would have exceeded
`
`the qualifications that the Petition has identified for the person of ordinary skill in
`
`the art. Ex. 2002 ¶ 29. As Dr. Wang explains, ordinary artisans in the field of
`
`speech recognition, many of whom he has taught and trained, do not have the
`
`expertise to design parallel computing systems of the type taught in Chen. Ex.
`
`2002 ¶ 29. Besides processors and memories, such a design would have to address
`
`core hardware issues such as registers, busses, controllers, communication
`
`protocols, and even cooling fan systems. Ex. 2002 ¶ 29. Such knowledge is not
`
`necessary or even particularly relevant to an ordinary artisan in the field of
`
`automatic speech recognition. Ex. 2002 ¶ 29. Indeed, even an eminent scholar in
`
`the field of automatic speech recognition such as Dr. Wang would not have the
`
`expertise to implement Chen’s teachings if asked to modify a personal computer to
`
`realize Chen’s parallel processing architecture. Ex. 2002 ¶ 29. Notably, parallel
`
`processing is a computer systems area that is distinct from the signal processing
`
`and machine learning area where speech recognition belongs. Ex. 2002 ¶ 29.
`
`
`
`
`
`
`
`
`
`- 12 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`Given that Jiang was expressly directed to “a general purpose computing
`
`device in the form of a conventional personal computer 20,” Ex. 1004, 5:4-7, an
`
`ordinary artisan would not have been motivated to substitute Jiang’s generic
`
`“processing unit 21”/”CPU 21” with the Cray supercomputer “parallel processing
`
`computer systems for performing . . . MIMD parallel processing” disclosed in
`
`Chen. Ex. 1005, 1:29-31; Ex. 2002 ¶ 30. By way of analogy, such a modification
`
`would have been akin to replacing a conventional engine in a conventional
`
`automobile with the engine from a Formula One car. Ex. 2002 ¶ 30. Although such
`
`a modification would appear on the surface to promise to make the conventional
`
`car faster, there is nothing obvious about it, and the potential motivation of “more
`
`speed” is not ultimately a motivation to make such a substitution because a
`
`Formula One engine is fundamentally unsuitable for use with a conventional
`
`commercially available automobile. Ex. 2002 ¶ 30.
`
`In the same vein, while replacing a conventional computer processor with a
`
`highly complex supercomputer processing architecture may superficially seem
`
`poised to make the conventional computer “faster,” an ordinary artisan would have
`
`known that such a substitution is anything but “obvious” and would make no
`
`sense. Ex. 2002 ¶ 30. Simply put, an ordinary artisan would not have been
`
`motivated to implement Jiang’s speech recognition teachings, which were intended
`
`
`
`
`
`
`
`
`
`- 13 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`for the computing environment of a generic personal computer, using Chen’s
`
`eight-processor, four-memory, multiple-instruction-multiple-data parallel
`
`processor. Ex. 2002 ¶ 30. Such a combination would have been the work of a
`
`highly unusual artisan using highly unordinary creativity, not the work of an
`
`ordinary artisan using ordinary creativity, as obviousness requires. Ex. 2002 ¶ 29.
`
`Moreover, an ordinary artisan would have understood that Chen’s parallel
`
`processing system is the opposite of Jiang’s CPU-based processing system. Ex.
`
`2002 ¶ 27. In particular, an ordinary artisan would have known that Jiang’s speech
`
`recognition technique is directed by a CPU, which operates in serial, and that
`
`Jiang’s teachings are thus not suited for implementation in Chen’s parallel
`
`processing environment. Id. Thus, the ordinary artisan would not have been
`
`motivated to implement Jiang’s CPU-based speech recognition teachings in Chen’s
`
`parallel processing system. Id.
`
`Further, as explained above, such modifications would have fundamentally
`
`altered the resulting device so that it would no longer be suitable as a “personal
`
`computer” due to the resulting cost and even size of the system. Ex. 2002 ¶ 31. The
`
`resulting system would have greatly limited the usability of speech recognition
`
`systems, including Jiang’s speech recognition techniques. Ex. 2002 ¶ 31. Thus, an
`
`ordinary artisan would not have been motivated to modify Jiang in order to use the
`
`
`
`
`
`
`
`
`
`- 14 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`hardware design of Chen. Ex. 2002 ¶ 31. In addition, for the reasons above, an
`
`ordinary artisan also would not have had a reasonable expectation of success in
`
`undertaking such a modification. Ex. 2002 ¶ 32.
`
`VII. The Petition fails to demonstrate obviousness as to a shared acoustic
`model memory containing acoustic model data for a cluster of processors
`In addition to limitation 1(a)’s requirement of “one or more clusters of
`
`processors, each of the one or more clusters of processors comprising: a plurality
`
`of processors;” limitation (b) further requires “an acoustic model memory storing
`
`acoustic model data,” and limitation (c) recites “wherein each of the plurality of
`
`processors is configured to compute a probability using the acoustic model data in
`
`the acoustic model memory.”
`
`Together, limitation 1(a)-(c) thus require that each of the plurality of
`
`processors in a cluster share an acoustic model storing acoustic model data. Ex.
`
`2002 ¶ 33.
`
`The specification of the ’319 patent describes that same architecture,
`
`explaining:
`
`“[T]he present invention provides a circuit in which speech
`recognition processing is performed in parallel by groups of
`processors operating in parallel in which each group accesses a
`common memory of lexical data. . . . Each processor within a
`group can access the same lexical data as any other processor in the
`group. The controller can thus control the parallel processing of
`input speech parameters in a more flexible manner. For example, it
`
`
`- 15 -
`
`
`
`
`
`
`
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`allows more than one processor to process input speech parameters
`using the same lexical data in a lexical memory. This is because the
`lexical data is segmented into domains which are accessible by
`multiple processors.” Ex. 1001, 3:44-58 (emphasis added).
`Figure 2 of the patent, annotated below, further illustrates that architecture
`
`by showing two groups of lexical tree processors, with each group containing
`
`multiple processors 1-k, and each group of processors connected to a dedicated
`
`“acoustic model memory.” Ex. 2002 ¶ 34, Ex. 1001, Fig. 2 (annotated).
`
`
`
`
`
`
`The Petition’s sole theory as to limitations 1(a)-(c) relies on the combination
`
`
`
`of Jiang and Chen. Pet. 14-31, 54. The Petition admits that Jiang does not teach the
`
`several processors to one acoustic model memory architecture of the challenged
`
`
`
`
`
`
`
`
`
`- 16 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`claims. Pet. 14. The Petition therefore again looks to Chen for the teaching of
`
`groups of processors placed into “separable clusters, each cluster having a common
`
`cluster shared memory that is symmetrically accessible by all of the processors in
`
`that cluster.” Pet. 23.
`
`According to the Petition’s theory, it would allegedly have been obvious to
`
`modify Jiang in view of Chen to create a system with multiple clusters of
`
`processors each with a shared memory. Pet. 24. In particular, the Petition proposes
`
`to “implement the speech recognition techniques taught by Jiang using the
`
`clustered processor computing platform of Chen to expedite the search process by
`
`leveraging shared memory among clustered processors.” Pet. 24. The Petition
`
`additionally asserts that “a POSITA would have stored at least a portion of the
`
`acoustic model data (HMM acoustic model from Jiang’s model memory 72) in the
`
`shared cluster memories 104a-d of each of the clusters 100a-d taught by Chen for
`
`access by processors 102a-d during speech recognition[.]” Pet. 24. The Petition
`
`later states, “in the proposed modified system, each of Chen’s processors 102a-d in
`
`respective clusters 100a-d would have computed probability scores as taught by
`
`Jiang using the acoustic model data stored in shared cluster memories 104a-d of
`
`each of the clusters 100a-d[.]” Pet. 30.
`
`
`
`
`
`
`
`
`
`- 17 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`The Petition, however, provides no reason why the ordinary artisan would
`
`have been motivated to specifically utilize each of Chen’s shared cluster memories
`
`104a-d as an “acoustic model memory” storing “acoustic model data,” as recited in
`
`limitations 1(a)-(c) of the challenged claims. Ex. 2002 ¶ 37.
`
`The Petition’s failure to provide any motivation for that crucial aspect of its
`
`combination theory is fatal. PersonalWeb Techs., LLC v. Apple, Inc., 848 F.3d 987,
`
`991 (Fed. Cir. 2017); In re NuVasive, Inc., 842 F.3d 1376, 1381-82 (Fed. Cir.
`
`2016); In re Warsaw Orthopedic, Inc., 832 F.3d 1327, 1333-34 (Fed. Cir.
`
`2016); Ariosa Diagnostics v. Verinata Health, Inc., 805 F.3d 1359, 1364-67 (Fed.
`
`Cir. 2015); ActiveVideo Networks v. Verizon, 694 F.3d 1312, 1328 (Fed. Cir.
`
`2012). When correctly applied, the requirement of a motivation to combine the
`
`prior art elements in the specific manner alleged “serves to prevent hindsight bias.”
`
`Medichem, S.A. v. Rolabo, S.L., 437 F.3d 1157, 1164 (Fed. Cir. 2006) (internal
`
`citations and quotations omitted).
`
`Notably, Chen does not contain any teachings regarding speech recognition
`
`systems, and Chen thus could not motivate using each of its cluster shared
`
`memories as acoustic model memories storing acoustic model data. Ex. 2002 ¶ 36,
`
`37. Jiang, on the other hand, does not teach clusters of processors with each cluster
`
`having its own dedicated acoustic model memory storing acoustic model data.
`
`
`
`
`
`
`
`
`
`- 18 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`Jiang thus likewise could not motivate the Petition’s specific design choice. Ex.
`
`2002 ¶ 37.
`
`Rather, the Petition’s convenient theory is plainly driven by hindsight, and
`
`simply reflects a reconstruction of the challenged claims arrived upon by starting
`
`with the claims themselves and working backwards. But “[i]n making obviousness
`
`determinations, the test is ‘whether the subject matter of the claimed inventions
`
`would have been obvious to one skilled in the art at the time the inventions were
`
`made, not what would be obvious to a judge after reading the patents in suit and
`
`hearing testimony.” Medichem, 437 F.3d at 1164 (emphasis original) (quoting
`
`Panduit Corp. v. Dennison Mfg. Co., 774 F.2d 1082, 1092 (Fed. Cir. 1985)).
`
`Because the Petition’s theory as to limitations 1(a)-(c) does not demonstrate
`
`why the artisan of ordinary skill “would have combined elements from specific
`
`references in the way the claimed invention does,” ActiveVideo, 694 F.3d at 1328
`
`(emphasis original), the Petition fails to demonstrate a reasonable likelihood of
`
`success, much less a high likelihood of success.
`
`VIII. The Petition fails to demonstrate obviousness as to “each of the plurality
`of processors [being] configured to compute a probability,” as recited in
`1(c)
`The Petition’s obviousness theory also fails at a second level. Whereas the
`
`Petition alleges it would have been obvious to combine Jiang and Chen such that
`
`each of Chen’s processors in a cluster of processors would have been configured to
`
`
`
`
`
`
`
`
`- 19 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`compute a probability, as recited in limitation 1(c), the Petition provides no actual
`
`reason why an ordinary artisan would have configured the combination of Jiang
`
`and Chen in that manner.
`
`Under the heading “Motivation to Combine Chen with Jiang for Claim
`
`1(c),” the Petition provides no reason whatsoever. Pet. 31. Instead, the Petition
`
`simply makes conclusory assertions as to what purportedly would have happened,
`
`without giving any motivation for why those assertions would have occurred. The
`
`entirety of the Petition’s “motivation” section for limitation 1(c) is below.
`
`
`
`Moreover, the Petition’s earlier statement that “[a] POSITA would have
`
`recognized it would have been advantageously efficient to use each of the clustered
`
`processors of Jiang-Chen to perform speech recognition,” Pet. 30, likewise says
`
`nothing at all. As the Federal Circuit has held, the “generic” assertion of
`
`“efficiency” is not evidence of a motivation to combine, as it “bears no relation to
`
`
`
`
`
`
`
`
`- 20 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`any specific combination of prior art elements” and “fails to explain why a person
`
`of ordinary skill would have combined elements from specific references in the
`
`way the claimed invention does.” ActiveVideo, 694 F.3d at 1328 (emphasis
`
`original). Furthermore, an ordinary artisan would know that even using each of
`
`Chen’s clustered processors for performing the speech recognition taught in Jiang
`
`would not necessarily lead to each of the processors being “configured to compute
`
`a probability,” as limitation 1(c) requires. Ex. 2002 ¶ 39. Rather, Jiang’s speech
`
`recognition technique requires various processing activities, and there is no
`
`particular reason why each of Chen’s multiple processors would have been
`
`dedicated to the specific task of computing a probability. Ex. 2002 ¶ 39.
`
`Likewise inadequate is the Petition’s assertion that “Chen teaches that for
`
`each cluster of processors, the plurality of processors in each cluster may perform
`
`processing in parallel, such that each processor of each of the clusters performs the
`
`intended operation.” Pet. 30 (emphasis removed). The fact that Chen’s processors
`
`can work in parallel on a given task is not a reason why they would have been
`
`specifically configured to each perform the particular task of computing a
`
`probability, as recited in limitation 1(c). Ex. 2002 ¶¶ 38-39.
`
`Accordingly, the Petition fails to demonstrate a reasonable likelihood of
`
`success as to limitation 1(c), much less a high likelihood of success.
`
`
`
`
`
`
`
`
`
`- 21 -
`
`
`
`
`
` IPR2023-00037
`
`PATENT OWNER’S PRELIMINARY RESPONSE
`
`IX. Conclusion
`For the reasons above, the Board should decline to institute the Petition.
`
`
`Respectfully submitted,
`
`/s/ Peter C. Knops
`
`Peter C. Knops, Reg. No. 37,659
`Kayvan B. Noroozi, Pro Hac Vice (to be filed)
`NOROOZI PC
`11601 Wilshire Blvd., Suite 2170
`Los Angeles, CA 90025
`Attorneys for the Patent Owner
`
`
`
`
`Date: March 15, 2023
`
`
`
`
`
`
`
`
`
`- 22 -
`
`
`
`
`
`CERTIFICATION OF WORD COUNT
`
`
`
`The undersigned hereby certifies that the portions of the above-captioned
`
`PATENT OWNER’S PRELIMINARY RESPONSE specified in 37 C.F.R. §
`
`42.24 have 4,175 words, in compliance with the 14,000 word limit set forth in 37
`
`C.F.R. § 42.24(b)(1). This word count was prepared using Microsoft Word 365.
`
`/Peter Knops/
`Peter Knops
`
`
`
`
`
`
`
`
`
`Counsel for Patent Owner
`
`
`
`
`
`
`
`
`
`Date: March 15, 2023
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`CERTIFICATION OF SERVICE
`
`
`
`The undersigned hereby certifies that the foregoing PATENT OWNER’S
`
`PRELIMINARY RESPONSE was served electronically via e-mail on March 15,
`
`2023, on the following counsel of record for Petitioner:
`
`Adam M. Sandwell
`Adam.Sandwell@eriseip.com
`
`
`
`
`
`
`Adam P. Seitz
`Adam.Seitz@eriseip.com
`PTAB@eriseip.com
`
`Jennifer C. Bailey
`Jennifer.Bailey@eriseip.com
`
`Erise IP, P.A.
`7015 College Blvd., Ste. 700
`Overland Park, Kansas 66211
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Date: March 15, 2023
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`/Peter C. Knops/
`Peter C. Knops
`NOROOZI PC
`11601 Wilshire Blvd., Suite 2170
`Los Angeles, CA 90025
`Tel.: 310-975-7074
`
`Counsel for Patent Owner
`
`
`
`