`
`Page 1
`
`·2· · · · · BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`·3
`
`·4· ·APPLE INC.,· · · · · · · )
`· · · · · · · · · · · · · · · )
`·5· · · · · · Petitioner,· · ·)
`· · · · · · · · · · · · · · · ) Case No. IPR2023-00033
`·6· · ·vs.· · · · · · · · · · )
`· · · · · · · · · · · · · · · ) U.S. Patent No. 7,587,319
`·7· ·ZENTIAN LIMITED,· · · · ·)
`· · · · · · · · · · · · · · · )
`·8· · · · · · Patent Owner.· ·)
`
`·9
`
`10
`
`11
`· · VIDEOCONFERENCE DEPOSITION OF DAVID V. ANDERSON, PH.D.
`12
`
`13· · · · · · · TAKEN ON BEHALF OF THE PETITIONER
`
`14
`· · · · · · · · · · · ·NOVEMBER 17, 2023
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22· Court Reporter:
`· · Lauren N. Lawrence, RPR, CCR-MO, CCR-KS
`23· Lexitas Legal
`· · 1608 Locust Street
`24· Kansas City, Missouri 64108
`· · 816.221.1160
`25· 800.280.3376
`
`IPR2023-00037
`Apple EX1035 Page 1
`
`
`
`·1· · · · · · · · · · · I N D E X
`
`Page 2
`
`·2· · WITNESS:· · · · · · · · · · · · · · · · · · PAGE
`
`·3· · DAVID V. ANDERSON, PH.D.
`
`·4· ·EXAMINATION BY MS. BAILEY· · · · · · · · · · · ·5
`
`·5
`
`·6
`
`·7· · · · · · ·(No exhibits were marked.)
`
`·8
`
`·9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`IPR2023-00037
`Apple EX1035 Page 2
`
`
`
`·1· · · · ·UNITED STATES PATENT AND TRADEMARK OFFICE
`
`Page 3
`
`·2· · · · · BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`·3
`
`·4· ·APPLE INC.,· · · · · · · · )
`· · · · · · · · · · · · · · · · )
`·5· · · · · · Petitioner,· · · ·)
`· · · · · · · · · · · · · · · · )· Case No. IPR2023-00033
`·6· · ·vs.· · · · · · · · · · · )
`· · · · · · · · · · · · · · · · )· U.S. Patent No. 7,587,319
`·7· ·ZENTIAN LIMITED,· · · · · ·)
`· · · · · · · · · · · · · · · · )
`·8· · · · · · Patent Owner.· · ·)
`
`·9
`
`10· · · · · · VIDEOCONFERENCE DEPOSITION OF DAVID V.
`
`11· ANDERSON, PH.D., produced, sworn, and examined on
`
`12· November 17, 2023, all parties appearing remotely,
`
`13· before Lauren N. Lawrence, RPR, CCR-MO, CCR-KS, in a
`
`14· certain cause now pending in the United States Patent
`
`15· and Trademark Office before the Patent Trial and
`
`16· Appeal Board, between APPLE INC., Petitioner, vs.
`
`17· ZENTIAN LIMITED, Patent Owner; on behalf of the
`
`18· Petitioner.
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`IPR2023-00037
`Apple EX1035 Page 3
`
`
`
`·1· · · · · · · · · ·A P P E A R A N C E S
`
`Page 4
`
`·2
`
`·3· APPEARING FOR PETITIONER:
`
`·4· · · ·Ms. Jennifer C. Bailey
`· · · · ·ERISE IP, P.A.
`·5· · · ·7015 College Boulevard
`· · · · ·Suite 700
`·6· · · ·Overland Park, Kansas 66211
`· · · · ·813.777.5600
`·7· · · ·jennifer.bailey@eriseip.com
`
`·8· · · ·Ms. Christina Canino, pro hac vice
`· · · · ·717 17th Street
`·9· · · ·Suite 1400
`· · · · ·Denver, Colorado 80202
`10· · · ·913.777.5600
`
`11
`· · APPEARING FOR PATENT OWNER:
`12
`· · · · ·Mr. Kayvan B. Noroozi
`13· · · ·NOROOZI PC
`· · · · ·11601 Wilshire Boulevard
`14· · · ·Suite 2170
`· · · · ·Los Angeles, California 90025
`15· · · ·310.975.7074
`· · · · ·kayvan@noroozipc.com
`16
`
`17· ALSO PRESENT:
`
`18· · · ·Adam Sandwell, Technical Analyst
`
`19
`
`20
`
`21
`
`22· Court Reporter:
`· · Lauren N. Lawrence, RPR, CCR-MO, CCR-KS
`23· Lexitas Legal
`· · 1608 Locust Street
`24· Kansas City, Missouri 64108
`· · 816.221.1160
`25· 800.280.3376
`
`IPR2023-00037
`Apple EX1035 Page 4
`
`
`
`·1· · · · · · IT IS HEREBY STIPULATED AND AGREED by and
`
`Page 5
`
`·2· between counsel for the Petitioner and counsel for the
`
`·3· Patent Owner that this deposition may be taken in
`
`·4· shorthand by Lauren N. Lawrence, RPR, CCR-MO, CCR-KS,
`
`·5· and afterwards transcribed into typewriting; and the
`
`·6· signature of the witness is expressly reserved.
`
`·7· · · · · · · · · ·*· · *· · *· · *· · *
`
`·8· · · · · · (Deposition commenced at 9:37 a.m.)
`
`·9· · · · · · · · ·DAVID V. ANDERSON, PH.D.,
`
`10· having been first duly sworn and examined on behalf of
`
`11· · · · · ·the Petitioner, testified as follows:
`
`12· · · · · · · · · · · · EXAMINATION
`
`13· BY MS. BAILEY:
`
`14· · · · ·Q.· ·Good morning, Dr. Anderson.· Can you --
`
`15· · · · ·A.· ·Good morning.
`
`16· · · · ·Q.· ·-- confirm for me, please, that you have
`
`17· electronic copies of the record for accessing today?
`
`18· · · · ·A.· ·Yes, I do.
`
`19· · · · ·Q.· ·Okay.· And could you please confirm that
`
`20· you don't have any notes, papers, or any other
`
`21· information regarding these matters in front of you
`
`22· other than the electronic copies of the record.
`
`23· · · · ·A.· ·I do have a list of some typos in front of
`
`24· me that I wanted to go over.
`
`25· · · · ·Q.· ·Okay.· Sure.
`
`IPR2023-00037
`Apple EX1035 Page 5
`
`
`
`Page 6
`·1· · · · ·A.· ·That's the only thing I have in front of me
`
`·2· that's not just the electronic copies.
`
`·3· · · · ·Q.· ·Sure.· Do you want to go ahead and go over
`
`·4· those typos now?
`
`·5· · · · ·A.· ·Sure.· In paragraph 140 -- I'm sorry.
`
`·6· Of -- in my 140 declaration, paragraph 4 --
`
`·7· · · · · · · (Reporter clarification.)
`
`·8· · · · ·A.· ·In my 140 declaration, paragraph 4, it says
`
`·9· "277."· That should say "140."
`
`10· · · · · · · In paragraph 12 of the same declaration, it
`
`11· says "3 years."· That should say "30 years."
`
`12· · · · · · · In paragraph 45 -- it's not a big deal, but
`
`13· there's an extra word.· "Have," h-a-v-e.
`
`14· · · · · · · And then in -- after paragraph 30, on
`
`15· page 15, the numbering restarts.· So when we've been
`
`16· discussing this, we give both the page number and the
`
`17· paragraph number because it restarts to 30 -- or I'm
`
`18· sorry to 23, which means that there are two 23s up to
`
`19· two 29s, and so we might need to be specific with
`
`20· those.
`
`21· · · · · · · And then in my 319 declaration,
`
`22· paragraph 4, "277" should say "319."
`
`23· · · · ·Q.· ·(By Ms. Bailey)· Thank you, Dr. Anderson.
`
`24· · · · · · · Was there any other further typographical
`
`25· errors you wanted to correct?
`
`IPR2023-00037
`Apple EX1035 Page 6
`
`
`
`·1· · · · ·A.· ·That's it.
`
`Page 7
`
`·2· · · · ·Q.· ·Do you understand that today's deposition
`
`·3· is covering two IPRs, one IPR for the 319 patent and
`
`·4· one IPR for the 140 patent?
`
`·5· · · · ·A.· ·Yes.
`
`·6· · · · ·Q.· ·So I'm going to ask you a number of
`
`·7· questions today.· And in an effort to avoid repetitive
`
`·8· questions that may be applicable to both IPRs, when I
`
`·9· ask a question, I'm going to try to make clear if it's
`
`10· specific to the 140 patent that I'm asking a question
`
`11· specific to the 140.· And, similarly, if it's specific
`
`12· to the 319, I'm asking a question specific for the 319
`
`13· patent.· If I don't specify, is it fair or are you
`
`14· comfortable with me not specifying and that you are
`
`15· understanding that your answer is applicable to both
`
`16· patents?
`
`17· · · · ·A.· ·Yes.
`
`18· · · · ·Q.· ·And, of course, if you have a question or
`
`19· you're unsure which patent we're talking about, please
`
`20· ask me, and I am very happy to specify.
`
`21· · · · ·A.· ·Okay.
`
`22· · · · ·Q.· ·Are you aware --
`
`23· · · · · · · MR. NOROOZI:· Just for a minute, I mean,
`
`24· I -- that raises a concern from my perspective,
`
`25· because there are different combinations of references
`
`IPR2023-00037
`Apple EX1035 Page 7
`
`
`
`Page 8
`·1· at issue, and there are different opinions and -- and
`
`·2· so forth.· So I think it raises, you know, a
`
`·3· significant risk of confusion if -- if we're talking
`
`·4· generically without specifying.· But I will just say
`
`·5· if, in any instance, there's a lack of ambiguity -- I
`
`·6· mean, there's a lack of clarity, then, you know, I'll
`
`·7· leave it to the witness and the questioning attorney
`
`·8· to sort that out between you.· I'll make my
`
`·9· objections.
`
`10· · · · · · · MS. BAILEY:· Understood.· And, Counsel, if
`
`11· you also are unclear which question or which patent I
`
`12· am asking a question for, please let me know.· Again,
`
`13· I'm happy to clarify.
`
`14· · · · · · · MR. NOROOZI:· Okay.
`
`15· · · · ·Q.· ·(By Ms. Bailey)· Dr. Anderson, are you
`
`16· aware of the legal concept in patent law regarding
`
`17· enablement of a patent?
`
`18· · · · ·A.· ·I'm not a lawyer, but I know that basically
`
`19· enablement has to do with whether the patent teaches
`
`20· how to implement or use a particular invention.
`
`21· · · · ·Q.· ·Do you think that the enablement standard
`
`22· requires a patent to lay out detail on how to make and
`
`23· use the claimed invention that would otherwise be
`
`24· known to the skilled artisan?
`
`25· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`IPR2023-00037
`Apple EX1035 Page 8
`
`
`
`Page 9
`·1· · · · ·A.· ·So I didn't understand that enablement was
`
`·2· at issue here, and so I haven't spent time preparing,
`
`·3· you know, to understand what exactly enablement means
`
`·4· in legal terms.
`
`·5· · · · ·Q.· ·(By Ms. Bailey)· Are you aware that the
`
`·6· claim priority date for the 140 and 319 patents is
`
`·7· February 4, 2002?
`
`·8· · · · · · · And if it's helpful for you, if you want to
`
`·9· turn to your declaration for the 140 patent,
`
`10· paragraph 14...
`
`11· · · · ·A.· ·Yes.· I see in paragraph 15, it says "I
`
`12· understand that the petition applies a priority date
`
`13· of February 4, 2002, for the challenged claims, and I
`
`14· apply the same date."
`
`15· · · · ·Q.· ·So if I refer to the priority date of the
`
`16· patents, can we agree that I'm referring to
`
`17· February 4, 2002?
`
`18· · · · ·A.· ·Yes.· Is that the same for the 319, though?
`
`19· Oh, it is.· I see it there.· Yes.· I'm good with that.
`
`20· · · · ·Q.· ·Could you please turn to your declaration
`
`21· for the 140 patent and read paragraph 16 to yourself?
`
`22· · · · ·A.· ·Okay.
`
`23· · · · ·Q.· ·It is -- is it your understanding that this
`
`24· paragraph is laying out the definition of a person of
`
`25· skill in the art as set forth in the petition?
`
`IPR2023-00037
`Apple EX1035 Page 9
`
`
`
`·1· · · · ·A.· ·Yes.
`
`Page 10
`
`·2· · · · ·Q.· ·Is it your understanding that this same
`
`·3· level of skill is used and applied for the 319 patent?
`
`·4· · · · ·A.· ·Yes.
`
`·5· · · · ·Q.· ·And do you understand that this level of
`
`·6· skill for a person of skill in the art includes both a
`
`·7· minimum education and a minimum years of experience?
`
`·8· · · · ·A.· ·Yes.
`
`·9· · · · ·Q.· ·Okay.· Do you dispute that this is a
`
`10· correct level of skill for the 140 and the 319
`
`11· patents?
`
`12· · · · ·A.· ·I do not.
`
`13· · · · ·Q.· ·Do you think you are a person of skill in
`
`14· the art applying this definition?
`
`15· · · · ·A.· ·Yes.
`
`16· · · · ·Q.· ·What are your qualifications that make you
`
`17· a person of skill in the art applying this definition?
`
`18· · · · ·A.· ·I have a master's degree in electrical and
`
`19· computer engineering and -- which I received in 1994.
`
`20· · · · · · · After that, I received a PhD in signal
`
`21· processing focusing on speech and audio processing,
`
`22· including speech recognition.
`
`23· · · · · · · After the PhD, I became a professor at
`
`24· Georgia Tech and have since done research and taught
`
`25· classes in speech recognition and other areas of
`
`IPR2023-00037
`Apple EX1035 Page 10
`
`
`
`·1· signal processing.
`
`Page 11
`
`·2· · · · ·Q.· ·Do you agree that this is a correct level
`
`·3· of skill for the 140 and 319 patents?
`
`·4· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`·5· · · · ·A.· ·I adopted this definition of a person of
`
`·6· ordinary skill in the art based on the petition's
`
`·7· definition.· I don't necessarily agree, but I'm --
`
`·8· it's a reasonable place to consider this from,
`
`·9· evaluate these patents from.
`
`10· · · · ·Q.· ·(By Ms. Bailey)· Why do you not necessarily
`
`11· agree with the definition?
`
`12· · · · ·A.· ·I haven't provided an alternative
`
`13· definition.· I didn't really think to.· I saw that
`
`14· this was a reasonable place to start, and since it was
`
`15· proposed by the petitioner, I adopted it.· If I were
`
`16· given time to consider, I might alter it in some way.
`
`17· I'm not sure yet.
`
`18· · · · ·Q.· ·Dr. Anderson, quite a bit of your opinions
`
`19· are based on what a skilled person would or would not
`
`20· know.· So are you stating today that you didn't really
`
`21· think to consider the level of skill?
`
`22· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`23· · · · ·A.· ·No.· That's not what I'm saying.
`
`24· · · · ·Q.· ·(By Ms. Bailey)· Could you please clarify
`
`25· then what you are saying?
`
`IPR2023-00037
`Apple EX1035 Page 11
`
`
`
`Page 12
`·1· · · · ·A.· ·I'm saying that the petitioner put forth a
`
`·2· person of ordinary skill in the art definition.· It
`
`·3· was reasonable.· Somebody with experience in speech
`
`·4· recognition.· I believe somebody with those particular
`
`·5· backgrounds could implement the patents at issue, and
`
`·6· so those are the ones that I used.· I took their
`
`·7· definition because that was where the arguments were
`
`·8· made from.· Didn't propose a different one.· That's
`
`·9· not my role.· Instead, I evaluated the obviousness and
`
`10· motivation to combine and other things from the
`
`11· definition given by the petitioner.· The main issue
`
`12· that I have -- as you said, a lot comes down to a
`
`13· person of ordinary skill in the art.
`
`14· · · · · · · The main problem I have isn't that this
`
`15· defined person of ordinary skill in the art could not
`
`16· make use of the claimed inventions in the 319 and 140
`
`17· patent.· The main issue is that the petitioner put
`
`18· forward some combinations where there are elements
`
`19· missing and would need to be filled in.· And this
`
`20· person of ordinary skill in the art would not have the
`
`21· requisite skills or understanding to do design in
`
`22· another field as required by the petitioner's
`
`23· combinations.
`
`24· · · · ·Q.· ·Do you have an opinion on what the
`
`25· appropriate level of skill should be for the 140 and
`
`IPR2023-00037
`Apple EX1035 Page 12
`
`
`
`·1· 319 patents?
`
`Page 13
`
`·2· · · · ·A.· ·As I've said, I adopted the one given, and
`
`·3· I think it's reasonable.· I'm not proposing anything
`
`·4· different.
`
`·5· · · · ·Q.· ·Do you think an engineer with a master's
`
`·6· degree in EE, CS, or CE would have been trained in
`
`·7· evaluating both the costs and benefits of a particular
`
`·8· design choice in developing a signal processing
`
`·9· architecture?
`
`10· · · · ·A.· ·Let me break that question into two parts.
`
`11· I do think that an electrical engineering student or
`
`12· CS, more particularly, in electrical engineering in
`
`13· computer engineering, would be very versed in
`
`14· evaluating design tradeoffs, in general, in whatever
`
`15· area of expertise is.· They would all have the -- be
`
`16· trained to evaluate design tradeoffs.· But in signal
`
`17· processing architecture, for example, it depends on
`
`18· what their background is.
`
`19· · · · · · · If, for example, you were in electrical
`
`20· engineering and trained in fiber optics or analog
`
`21· circuits, you would have a very different toolbox with
`
`22· which to evaluate design tradeoffs.· You would
`
`23· understand the need to, but the specifics would depend
`
`24· on the discipline.
`
`25· · · · ·Q.· ·And do you agree that the definition for
`
`IPR2023-00037
`Apple EX1035 Page 13
`
`
`
`Page 14
`·1· the skilled person provided in the petitions for the
`
`·2· 140 and 319 patents allows for two years of experience
`
`·3· after graduation?
`
`·4· · · · ·A.· ·Yes.
`
`·5· · · · ·Q.· ·Could there be a person that studied
`
`·6· high-performance computing and parallel processing in
`
`·7· their bachelor's degree, studying speech recognition
`
`·8· in their master's and then spent two years building a
`
`·9· system architecture specific to speech recognition,
`
`10· does such a person potentially exist?
`
`11· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`12· · · · ·A.· ·You're describing something that I wouldn't
`
`13· call a person of ordinary skill in the art, in general
`
`14· or specifically with this.· Someone who -- who studies
`
`15· high-performance computer architecture in their
`
`16· bachelor's degree would not have the knowledge and
`
`17· experience to design new high-performance architecture
`
`18· wholesale.· That usually comes at the graduate level.
`
`19· Furthermore, such a person would have focused on
`
`20· classes in circuit design, computer architecture, not
`
`21· in signal processing.
`
`22· · · · · · · At Georgia Tech, we have three
`
`23· undergraduate classes in signal processing for those
`
`24· desiring to take that chain, plus classes in signal
`
`25· processing applications that prepare somebody for
`
`IPR2023-00037
`Apple EX1035 Page 14
`
`
`
`Page 15
`·1· graduate study in signal processing.· If you were to
`
`·2· go straight into that, you would need to go back and
`
`·3· take the prerequisite courses to understand just the
`
`·4· signaling processing concepts.
`
`·5· · · · · · · Once that's done, they would then study
`
`·6· speech recognition.· In a master's degree, by the time
`
`·7· they got the prerequisite chain out of the way, they
`
`·8· should have already graduated.· So you're -- you're
`
`·9· giving some hypothetical superman that just doesn't
`
`10· fit the reality of a ord- -- person of ordinary skill
`
`11· in the art.
`
`12· · · · ·Q.· ·(By Ms. Bailey)· As of the priority date of
`
`13· the 140 and 319 patents, were persons working on
`
`14· speech recognition?
`
`15· · · · ·A.· ·In general, were there people working on
`
`16· speech recognition as of those priority dates?· Is
`
`17· that what you're asking?
`
`18· · · · ·Q.· ·Correct.· That's my question.
`
`19· · · · ·A.· ·Yes.
`
`20· · · · ·Q.· ·Prior to February 2002, in the field of
`
`21· speech recognition, were there methods available to
`
`22· make speech recognition more efficient.
`
`23· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`24· · · · ·A.· ·In general, yes.
`
`25· · · · ·Q.· ·(By Ms. Bailey)· Did these methods include
`
`IPR2023-00037
`Apple EX1035 Page 15
`
`
`
`·1· developments and hardware architecture for speech
`
`Page 16
`
`·2· recognition?
`
`·3· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`·4· · · · ·A.· ·Most of the work -- I would say the vast
`
`·5· majority of the work in speech recognition was on
`
`·6· software systems and the mathematics behind them.
`
`·7· There are a few groups that focus also on hardware
`
`·8· accelerators.· Most of those were cited either in the
`
`·9· specifications of the patents or in the prior art, and
`
`10· those generally involved multi-disciplinary teams or
`
`11· people who had specific training in hardware
`
`12· architectures and speech recognition applications.
`
`13· · · · ·Q.· ·(By Ms. Bailey)· As of the priority date of
`
`14· the 140 and 319 patents, was the use of parallel
`
`15· processing techniques known to improve speech
`
`16· recognition?
`
`17· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`18· · · · ·A.· ·Based on the prior art that's been
`
`19· presented, some -- actually, it wasn't known generally
`
`20· to improve speech recognition.· There were some cases
`
`21· where parallel processing was known, but there are
`
`22· also cases -- this is a very complex problem, for
`
`23· example cited by Mathew, where parallel lies
`
`24· [phonetic] in the search stage was shown to actually
`
`25· decrease the overall performance due to the
`
`IPR2023-00037
`Apple EX1035 Page 16
`
`
`
`·1· synchronization requirements and handling and
`
`Page 17
`
`·2· coordinating the data along the different modules.
`
`·3· · · · ·Q.· ·(By Ms. Bailey)· You just prefaced the
`
`·4· answer to my previous question as based on the prior
`
`·5· art that's been presented, but I am asking you, as a
`
`·6· person of skill in the art, per your earlier
`
`·7· testimony, whether you knew that adding parallel
`
`·8· processing increased the efficiency for speech
`
`·9· recognition?
`
`10· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`11· · · · ·A.· ·So -- so I'm clear, you're saying as a
`
`12· person of ordinary skill in the art, in 2002, in that
`
`13· situation, would I have known that parallel processing
`
`14· would increase the efficiency of speech recognition?
`
`15· Is that -- is that your question?
`
`16· · · · ·Q.· ·(By Ms. Bailey)· Let me rephrase my
`
`17· question so it's clear and so the record is clear.
`
`18· · · · · · · As of 2002, are you aware of research being
`
`19· performed to increase the efficiency of speech
`
`20· recognition by implementing parallel processing
`
`21· methods?
`
`22· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`23· · · · ·A.· ·So in 2002, I don't recall what I knew at
`
`24· that time, but as I'm sitting here now and looking
`
`25· back at references, citations and other things about
`
`IPR2023-00037
`Apple EX1035 Page 17
`
`
`
`·1· the prior art in 2002, there are multiple teachings
`
`Page 18
`
`·2· about parallelization.· Parallelization in speech
`
`·3· recognition is a huge topic.· It's like saying "Is
`
`·4· medication known to help a disease?"· What medication?
`
`·5· What disease?
`
`·6· · · · · · · In speech recognition, you have multiple
`
`·7· phases or multiple stages or multiple steps, and most
`
`·8· of the prior art that did -- attempted parallelization
`
`·9· with signal processing worked on parallelization, the
`
`10· operation of those steps.· For example, having feature
`
`11· extraction running in parallel with distance
`
`12· calculation or something like that.· I don't recall
`
`13· seeing anything about -- no.· I guess there is
`
`14· something.
`
`15· · · · · · · There are a variety of parallelization
`
`16· techniques.· And as I quoted from Mathew, one of the
`
`17· prior art references earlier, people were exploring
`
`18· parallelization in many aspects of the speech
`
`19· recognition problem.· Mathew specifically looked at
`
`20· parallelizing the search stage, which I assume is at
`
`21· the bottom of your question, because that's what these
`
`22· patents are associated with.
`
`23· · · · · · · Mathew found that their implementation --
`
`24· and they were experts in the area -- actually did not
`
`25· lead to an improvement in performance.· So there was
`
`IPR2023-00037
`Apple EX1035 Page 18
`
`
`
`·1· work in parallelization.· Sometimes it paid off.
`
`Page 19
`
`·2· Sometimes it didn't.· It very much depended on the
`
`·3· particular problem and the implementation.
`
`·4· · · · ·Q.· ·(By Ms. Bailey)· As of the priority date of
`
`·5· the 140 and 319 patents, was there a desire for
`
`·6· improving speech recognition systems by making them
`
`·7· more flexible?
`
`·8· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`·9· · · · ·A.· ·What do you mean by "more flexible"?
`
`10· · · · ·Q.· ·(By Ms. Bailey)· Please turn to the 140
`
`11· patent, column 4, line 6.· Feel free to read any
`
`12· portions surrounding that section that you'd like to
`
`13· read to yourself to familiarize yourself.
`
`14· · · · ·A.· ·Oh, I was reading the wrong column.· Sorry.
`
`15· · · · ·Q.· ·It would probably be helpful to you just to
`
`16· read the entire paragraph that begins on column 3,
`
`17· line 62 and extends to column 4, line 18.
`
`18· · · · ·A.· ·Thank you.
`
`19· · · · · · · Okay.
`
`20· · · · ·Q.· ·Do you see that this paragraph that you
`
`21· just read in the 140 patent mentions "flexibility" at
`
`22· least three times?
`
`23· · · · ·A.· ·Yes.
`
`24· · · · ·Q.· ·So as of the priority date of the 140 and
`
`25· 319 patents, was there a desire for improving speech
`
`IPR2023-00037
`Apple EX1035 Page 19
`
`
`
`·1· recognition systems by making them more flexible?
`
`Page 20
`
`·2· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`·3· · · · ·A.· ·So your question is very general.
`
`·4· "Flexible" could mean so many different things.· Many,
`
`·5· many different things.· Flexible in terms of
`
`·6· understanding different languages.· That would be
`
`·7· flexibility.· Flexible in terms of being easily
`
`·8· tuneable to operate under different noise situations.
`
`·9· Flexible in that it was software only, so that it
`
`10· could be implemented in a variety of different
`
`11· platforms using only a C compiler.
`
`12· · · · · · · This particular flexibility talks about
`
`13· something that I haven't seen expressed as a goal in
`
`14· any of the prior art systems.· It really hasn't come
`
`15· up because this one talks about the benefits of a
`
`16· limited segmentation of the lexical data while
`
`17· maintaining a flexibility in how the tasks are divided
`
`18· up among that group.· And since we didn't have
`
`19· anything like that in the prior art, nobody has
`
`20· expressed that that was a lack.
`
`21· · · · ·Q.· ·(By Ms. Bailey)· Do you think that it was
`
`22· known as of the priority date of the 140 patent and
`
`23· 319 patent that by having a plurality of processors in
`
`24· a group with a common memory there would be increased
`
`25· flexibility in the processing?
`
`IPR2023-00037
`Apple EX1035 Page 20
`
`
`
`·1· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`Page 21
`
`·2· · · · ·A.· ·So let's see.· I'm trying to parse your
`
`·3· question here.· If you have multiple processors in the
`
`·4· common memory, then there are more things to consider
`
`·5· how you would divide up the tasks, for example; how
`
`·6· you would schedule things; whether you're going to use
`
`·7· threading.
`
`·8· · · · · · · Generally, I remember what it was like to
`
`·9· develop multithreaded applications in -- at that time.
`
`10· And multithreading libraries and techniques were still
`
`11· under continual development.· So there is -- there are
`
`12· more design choices that could be made.· And if that's
`
`13· what you mean by "flexibility," then yes.· That's not
`
`14· the same flexibility as -- as is discussed in 4 -- in
`
`15· column 4 of the 140 patent.
`
`16· · · · ·Q.· ·(By Ms. Bailey)· As of the priority date
`
`17· for the 140 and 319 patents, was there a general
`
`18· desire to make speech recognition systems scalable to
`
`19· handle larger vocabularies?
`
`20· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`21· · · · ·A.· ·At the time of the 140 patent, very large
`
`22· vocabulary speech recognition was known.· And at the
`
`23· time, once you are at a very large vocabulary of
`
`24· speech recognition, the focus is more on improving the
`
`25· performance of the recognizer as a signal processing
`
`IPR2023-00037
`Apple EX1035 Page 21
`
`
`
`·1· task.· There is one other aspect of this.· And that
`
`Page 22
`
`·2· was at the time also some of the commercial speech
`
`·3· recognition approaches and others were focused on
`
`·4· domain-specific, large-vocabulary speech recognition.
`
`·5· · · · · · · That's something that Thelen did.· Thelen
`
`·6· solved that problem by just running many different
`
`·7· speech recognizers each specific to that domain.
`
`·8· · · · · · · But even now, there's -- we don't have the
`
`·9· concept of extremely super large speech recognition
`
`10· systems.· So the concept of a large vocabulary speech
`
`11· recognition system was known, and that was what the
`
`12· 140 patent was addressing.
`
`13· · · · ·Q.· ·(By Ms. Bailey)· Are you aware of a concept
`
`14· of reduction of practice in patent law?
`
`15· · · · · · · MR. NOROOZI:· Objection to form.
`
`16· · · · ·A.· ·I know there is such a concept.· I couldn't
`
`17· opine on what meets that.
`
`18· · · · ·Q.· ·(By Ms. Bailey)· Do you think it would have
`
`19· been difficult for a skilled person to have made the
`
`20· claimed invention of the 140 and 319 patents?
`
`21· · · · · · · MR. NOROOZI:· Objection.· Form.
`
`22· · · · ·A.· ·I haven't formed an opinion on that. I
`
`23· think, generally, that a person of ordinary skill in
`
`24· the art and speech recognition as described could make
`
`25· use of the 140 patent.· But since I don't know what
`
`IPR2023-00037
`Apple EX1035 Page 22
`
`
`
`·1· "reduction of practice" means specifically, I mean,
`
`Page 23
`
`·2· I'm not ready to give a formal opinion on it, but I
`
`·3· was satisfied that there is sufficient detail in the
`
`·4· 140 and 319 patents that it could be made use of by
`
`·5· someone of ordinary skill in the art.
`
`·6· · · · ·Q.· ·(By Ms. Bailey)· Would coordinating
`
`·7· multiple caches be something that would need to be
`
`·8· addressed when implementing the invention -- the
`
`·9· claimed inventions of the 140 and 319 patents?
`
`10· · · · ·A.· ·So the 140 and 319 specifically talk about
`
`11· how the memory structure is to work.· It gives express
`
`12· details on the different memories on the buses, on the
`
`13· fact that you will have local memories at the
`
`14· processors for control, and it says what information
`
`15· is stored where.
`
`16· · · · · · · In my declaration, I discuss the difficulty
`
`17· of various things pertaining to designing a -- a
`
`18· moving -- for example, Jiang's recognition system or
`
`19· techniques to Chen's clustered parallel processing
`
`20· system.· And coordinating multiple caches is one of
`
`21· the things that's mentioned.· It turns out that the
`
`22· 140 has solved that.· But without the benefit of the
`
`23· 140 patent, this is a question that would have to be
`
`24· addressed by anyone who was to try to move Jiang's
`
`25· system onto Chen's architecture.
`
`IPR2023-00037
`Apple EX1035 Page 23
`
`
`
`Page 24
`·1· · · · ·Q.· ·Can you point me to where in the 140 patent
`
`·2· that there's a discussion of solving the problem of
`
`·3· caches and storage that you just addressed?
`
`·4· · · · ·A.· ·Yes.· This will take me a minute because I
`
`·5· don't have my marked-up copies.
`
`·6· · · · ·Q.· ·Understood.· And feel free to word search
`
`·7· if that will make it faster for you.
`
`·8· · · · ·A.· ·So for starters -- and I'll feed you stuff
`
`·9· as I run into it, so there might be pauses in between.
`
`10· · · · · · · The 140 patent and the specification for
`
`11· the 319 is the same.· So we'll just go through it -- a
`
`12· description of the 140 patent.
`
`13· · · · · · · The 140 patent specifies what program
`
`14· data -- program and data memory is, where the language
`
`15· model memory is, where the processor is, where the
`
`16· results memory are.· It also specifies where the
`
`17· feature verdict buffer is, and it specifies connection
`
`18· for each of these to their respective processors that
`
`19· are needed.
`
`20· · · · · · · It expresses where the acoustic model
`
`21· memory is.· The acoustic model memory is specified to
`
`22· hold the acoustic model, which would be a fixed thing.
`
`23· · · · · · · Finally, it gives lexical tree processors
`
`24· connected to acoustic model memory, the results
`
`25· memory, language model processor, and the feature
`
`IPR2023-00037
`Apple EX1035 Page 24
`
`
`
`·1· verdict buffer, and so this is a different
`
`Page 25
`
`·2· architecture than Chen.· Chen, instead, does not have
`
`·3· a universally connected results memory.
`
`·4· · · · · · · Chen -- and this is particularly relevant
`
`·5· for the 319 -- does not have a universally connected
`
`·6· feature verdict buffer.· Chen, instead, envisions
`
`·7· having distributed control and data which is to be
`
`·8· shared with -- globally has to be hopped from unit to
`
`·9· unit.
`
`10· · · · · · · In such a case, the data will be updating
`
`11· in the local memories -- in the cluster memories
`
`12· accessible to all processors because that's the only
`
`13· memory he discloses having, and that's the memory that
`
`14· Schmandt has pointed to as fulfilling all the
`
`15· operations in the combination.
`
`16· · · · · · · So by having this custom architecture in
`
`17· the 140 patent, you don't have to worry about cache
`
`18· consistency with the acoustic model memory because the
`
`19· acoustic model is established at the beginning.
`
`20· · · · · · · On the other hand, when you have the
`
`21· cluster memory continually being updated as items are
`
`22· being cascaded through and it's being used for
`
`23· multiple purposes, then cache consistency becomes an
`
`24· issue.
`
`25· · · · · · · Okay.· So now let's look at other things
`
`IPR2023-00037
`Apple EX1035 Page 25
`
`
`
`·1· relevant to your question.
`
`Page 26
`
`·2· · · · ·Q.· ·And, Dr. Anderson, before you move on, for
`
`·3· clarification, could you tell me where you were
`
`·4· looking at in the 140 patent in answering that
`
`·5· question?· Was there something specific?
`
`·6· · · · ·A.· ·I started on Figure 2.
`
`·7· · · · ·Q.· ·Thank you.
`
`·8· · · · ·A.· ·So the memory structure that I've been
`
`·9· giving the overview of, based on Figure 2, the
`
`10· description starts at the bottom of column 1.· And
`
`11· through column 4, it gives an overview of the
`
`12· arrangement.
`
`13· · · · · · · And then the specifics are really starting
`
`14· more in column 5, where it talks about the various
`
`15· memories and their uses.· And even, for example, in
`
`16· column 6, it specifies specific items that are to be
`
`17· stored in the memory and how they are to be accessed
`
`18· and by whom.· And...
`
`19· · · · · · · The on-board memory for the various
`
`20· processors is described, for example -- it's
`
`21· elsewhere, but it's mentioned in column 9, about
`
`22· halfway down, around line 43.
`
`23· · · · · · · By having local -- or having on-board
`
`24· memory for the processor in addition to the shared
`
`25· memory, you then avoid cache inconsistency problems
`
`IPR2023-00037
`Apple EX1035 Page 26
`
`
`
`·1· because things that you're operating on are now
`
`Page 27
`
`·2· isolated to specific processors with their local
`
`·3· caches.
`
`·4· · · · ·Q.· ·Thank you, Dr. Anderson.· You just gave a
`
`·5· lot of information about what was in the 140
`
`·6· specification.· Could you please turn to Claim 1 of
`
`·7· the 140 patent?
`
`·8· · · · ·A.· ·Okay.
`
`·9