`
`
`571-272-7822
`
`
`
`
`
`
`
`
`
`
`
`Case IPR2015-00343 Paper No. 29
`Case IPR2015-00345 Paper No. 29
`Case IPR2015-00347 Paper No. 29
`Case IPR2015-00348 Paper No. 29
`April 13, 2016
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`____________
`
`GOOGLE, INC.,
`Petitioner,
`
`v.
`
`NETWORK-1 TECHNOLOGIES, INC.,
`Patent Owner.
`____________
`
`Case IPR2015-00343 (Patent 8,640,179 B1)
`Case IPR2015-00345 (Patent 8,205,237 B2)
`Case IPR2015-00347 (Patent 8,010,988 B2)
`Case IPR2015-00348 (Patent 8,656,441 B1)
`____________
`
`Held: March 9, 2016
`____________
`
`
`
`
`
`BEFORE: KEVIN F. TURNER, LYNNE E. PETTIGREW, and
`JON B. TORNQUIST, Administrative Patent Judges.
`
`The above-entitled matter came on for hearing on Wednesday,
`March 9, 2016, commencing at 2:05 p.m., at the U.S. Patent and
`Trademark Office, 600 Dulany Street, Alexandria, Virginia.
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`
`
`APPEARANCES:
`
`ON BEHALF OF THE PETITIONER:
`
`
`
`
`
`
`
`
`
`
`JAMES J. ELACQUA, ESQ.
`DOUGLAS R. NEMEC, ESQ.
`ANDREW GISH, ESQ.
`Skadden, Arps, Slate, Meagher & Flom LLP
`525 University Avenue
`Palo Alto, California 94301
`
`
`
`
`
`
`
`
`and
`
`RICHARD A. SONNENTAG, ESQ.
`Google, Inc.
`1600 Ampitheatre Parkway
`Mountain View, California 94043
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`GREGORY DOVEL, ESQ.
`Dovel & Luner LLP
`201 Santa Monica Boulevard, Suite 600
`Santa Monica, California 90401
`
`and
`
`CHARLES R. MACEDO, ESQ.
`Amster Rothstein & Ebenstein LLP
`90 Park Avenue
`New York, New York 10016
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ON BEHALF OF THE PATENT OWNER:
`
` 2
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`
`P R O C E E D I N G S
`- - - - -
`JUDGE PETTIGREW: Good afternoon, everyone.
`This is a consolidated hearing for four cases, IPR2015-00343,
`IPR2015-00345, IPR2015-00347, and IPR2015-00348, Google,
`Inc. versus Network-1 Technologies, Inc.
`Each side has 60 minutes to argue. Petitioner, you have
`the ultimate burden of establishing unpatentability, so you will
`argue first. Patent Owner then will present its opposing
`argument. Finally, Petitioner may use any time it has reserved for
`rebuttal to respond to Patent Owner's argument.
`Judge Turner is joining us by video and audio from our
`Silicon Valley office, and won't have the benefit the visual cues
`in the room. So, when you speak about an exhibit or a
`demonstrative, please begin by identifying it with specificity,
`including the particular page or slide number. Also, please be
`sure to speak into the microphone to ensure that Judge Turner can
`hear you.
`Counsel, when you begin your argument, please
`identify yourself and the party you represent for the record.
`Petitioner, you may begin when ready.
`MR. ELACQUA: Thank you, Your Honor.
`
` 3
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`JUDGE PETTIGREW: And do you wish to reserve any
`rebuttal time?
`MR. ELACQUA: Yes, we wish to reserve 35 minutes.
`JUDGE PETTIGREW: Thirty-five minutes?
`MR. ELACQUA: Yes.
`JUDGE PETTIGREW: Thank you.
`MR. ELACQUA: And we have hard copies of the
`slides for the Board, if you would like.
`JUDGE PETTIGREW: Yes, that would be useful,
`thank you.
`JUDGE TURNER: And, Judge Pettigrew, can I just
`confirm that you can hear me fine?
`JUDGE PETTIGREW: We can hear you, Judge
`Turner, thank you.
`JUDGE TURNER: And I apologize to the parties for
`the delay, which perhaps was on our end out here in Silicon
`Valley. So, accept my apologies for our brief delay in getting
`started today.
`JUDGE PETTIGREW: All right, please begin.
`MR. ELACQUA: Thank you. Good afternoon, Your
`Honors. My name is Jim Elacqua and I'm hear this afternoon
`with Doug Nemec and Andrew Gish, also from Skadden, and
`Rich Sonnentag, in-house counsel for Google. We are here today
`representing Google on these four IPRs that involve 90 claims.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 4
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`And this afternoon, I'm going to be using a slide deck, and the
`slide deck is Exhibit 1022 for IPRs 343, 347 and 348, and Exhibit
`1021 for IPR 345. The numbering of the slides is identical for all
`the exhibits.
`Now, even though we have four IPRs and 90 claims,
`there are actually a limited number of issues for the Board to
`decide, and also interesting, the dependent claims are not
`contested for any of the patents and the motivation to combine is
`not contested on any of the obviousness issues. So, that again is
`limiting the number of decisions that the Board is needing to
`make.
`
`There are only really two main limitations contested,
`and if we go to slide 2, we will see that there are three references
`that we are going to be dealing with this afternoon, Ghias,
`Iwamura and Conwell, and the two main limitations are
`nonexhaustive search, regarding all the patents and all the
`references, and the near neighbor terms, which include
`identifying a neighbor and neighbor search. And that involves
`three of the patents, the '988, '179 and '441, and all of the
`references.
`There's another limitation we'll talk about and that's
`determining an action, which only involves the '988 patent and
`the Ghias reference. Now, if we go to slide 3, we will see that all
`of the contested limitations were widely known, and we took the
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 5
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`deposition of Dr. Karypis, and Dr. Karypis was at deposition, it
`was offered as an expert by Network-1, and we'll refer to his
`deposition from time to time. But in his deposition, we asked
`him, did Dr. Cox in the disclosure of his patents purport to have
`invented nonexhaustive searching, and he said no. And we also
`asked him if it was known in the art, and he said that that concept
`was known in the art.
`We also similarly asked him about neighbor searching,
`did Dr. Cox invent that, and he said, I don't believe so. And we
`asked if that concept was well known and he agreed that that
`concept was well known.
`So, we're dealing with what these limitations are, they're
`very broadly defined in these claims, and they're also very well
`known in the art. First I would like to discuss the Ghias
`reference, I would like to go reference by reference, I think it's
`easiest that way. And the Ghias reference involves, excuse me,
`the issues, if we go to slide 6, nonexhaustive, identifying a
`neighbor and determining an action. And the Board has actually
`provided us with a claim construction on this issue.
`Next slide. We'll go to slide 6 now, and we will see that
`the claim construction that the Board offered was a search that
`locates a match without a comparison of all possible matches, and
`we'll also see that in the Board's '988 institution decision, the
`Board agreed that anything that was not brute force, any search
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 6
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`that was not considered to be a brute force search, is a
`nonexhaustive search.
`Now, Ghias, the question is, does Ghias teach a
`nonexhaustive search, and Ghias does. If we go to slide 7, we
`will see that Ghias self-identifies a different than brute force
`search, and the Board has properly identified a second search in
`Ghias that also is a nonexhaustive search. So, let's deal with them
`one at a time, quickly.
`The first one is, Ghias, just to give some background,
`represents the queries and the references as a string of characters,
`U, D and S, which represent pitch differences between the notes,
`because after all, we're looking for a musical work. Now, Ghias
`matches a short query to certain melody segments within the
`song, and Ghias actually describes its comparison of the query to
`each reference as nonexhaustive.
`So, let's look at slide 8, and we will see how it does that.
`Ghias tells us, in Exhibit 1010, slide 8, that running times have
`ranged from the order of mn for brute force algorithm to the order
`of kn or the order of nlog(m). Now, what that tells us is what
`Ghias is saying is that what they're using for a search is
`something different than brute force, and as I said, the Board
`recognized and the parties are not disputing that searches that are
`different from brute force are nonexhaustive searches.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`
` 7
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`JUDGE PETTIGREW: Is that the comparison of
`excerpts from the melody to a single record that's being discussed
`in column 6?
`MR. ELACQUA: No, that was brought up in a
`different context, but this is the same quote that that issue has
`arisen from. And you're correct, Your Honor, that what they have
`asked here in their office -- in their Patent Owner response, they
`made the statement that this was restricted to a single song, and
`we can deal with that now or we can wait, but, Your Honor, our
`position is that this is not restricted. This is a definition that they
`provided that they're saying the searches that they're using, they're
`using searches other than brute force.
`So, for purposes of our prima facie case here, this
`shows that Ghias actually is using a nonexhaustive search. And
`we do not believe it's limited to a single song. And, in fact, as we
`can show, Ghias actually demonstrates, also, and teaches, that
`you can pack all of the songs into a single file. And to the extent
`that you can pack them all in a single file, you could use that
`same nonexhaustive search.
`Now, we also looked to its second reason why Ghias is
`teaching us a nonexhaustive search and this is one that the board
`found as well, and that is in looking at Ghias, we see that it is
`dealing with a multiple search process, where subsequent
`searches consider progressively narrower sets of references.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 8
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`So, looking at slide 5 for a moment, we'll see that this is
`the way that it actually works in Ghias, with each subsequent
`search looking at a more restricted list. Now, if this list becomes
`too large, what Ghias tells us is that the user can perform a new
`query on the restricted search list, which consists of just the songs
`that were retrieved. So, looking, again, if we look at -- we can
`see that Ghias says this on slide 10.
`Now, looking again, if we look at slide 9, we'll see that
`this restricted list with the new query, the fact is, is that you're not
`considering all possible matches, which is the requirement in the
`claim construction that the Board provided. So, Ghias actually
`provides a nonexhaustive search. And the Board agreed, on slide
`11, we'll see that in the Board's '988 institution decision, the
`Board actually agreed that with this new query on the restricted
`search list, the Board goes on to say, Ghias makes clear that the
`search need not be exhaustive.
`So, again, we presented two reasons, it's self-identifying
`in the way Ghias has presented this as a non-brute force search,
`and also just by the way that the search actually operates.
`I'd like to move on to the second issue for Ghias, which
`is the neighbor search, and the Board has also provided us with a
`definition for neighbor search, which is identifying a neighbor as
`identifying a close but not necessarily exact or closest match.
`That's on slide 12.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 9
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`Now, the parties have different positions on this as to
`what a neighbor search includes. Google takes the position that it
`includes searches that always find the closest match. The Patent
`Owner has a new interpretation that they set forth in their
`response that said, oh, no, that wording of the claim construction
`does not include searches that always find the closest match. So,
`with that said, if we go, you can see it in this diagram, those
`searches are called nearest neighbor searches. Our position is that
`they are included in the definition of neighbor search, Patent
`Owner's position is that they are not.
`And there are several reasons why that is wrong. The
`first one is, just as a matter of grammar, if we look at the
`grammar on slide 14, we see that the Board's language reads, "not
`necessarily exact or closest." What Network-1 would have us
`read this to mean is "necessarily not exact or closest." And that is
`not a fair grammatical reading of the words themselves.
`But just as important, we have a contradiction with the
`intrinsic record. The intrinsic record contradicts by the claim
`language itself that it improperly excludes searches that always
`find the closest match, which are nearest neighbor searches. So,
`if we look at slide 15, we will see, in claim 15 of the '988 patent,
`it calls for identification based on a nonexhaustive search
`identifying a neighbor.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`
` 10
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`Now, if we go to slide 16, we will see that there are
`three dependent claims, 18, 19 and 20, that both tell us that it is
`based on these new types of searches, kd-trees, vantage point
`trees and middle vantage point trees. But the specification in the
`Cox patents tell us that these particular types of searches are
`nearest neighbor searches. In fact, the '237 patent, which is
`Exhibit 1001, at column 9, lines 7 to 9, tells us that kd-tree
`method, which finds the nearest neighbor with certainty, and it's
`the same for all of these that are listed in these dependent claims.
`Why is that important? Because the fact that it's
`identifying these as a nearest neighbor search, the Cox patents tell
`us -- can we see slide 17, please? The Cox patents tell us, Exhibit
`1021, that the nearest neighbor search always finds the closest
`point to the query. So, the nearest neighbor searches are a type of
`neighbor search, just by a plain reading of the claim language
`itself.
`
`So, Petitioner's interpretation that it excludes finding the
`closest -- always finding the closest match is actually incorrect,
`because the claims call for that specifically.
`Now, there could be no dispute that the Ghias reference
`actually locates a close match. If we look at the slide 18, we will
`see that right within the reference itself, it talks about the output
`as a rank list of approximately matching melodies. It tells us.
`These are close matches. They're approximately matching
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 11
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`melodies. And, of course, figure 1 on slide 18 confirms that and
`shows that the ranked list is part of the output.
`There can be no dispute that there is a close match here.
`And a close match is what it takes to meet the claim language and
`the claim construction that the Board has provided.
`There's one other in Ghias, and that is determining an
`action. Now, I can't believe that this is going to be seriously
`challenged, but I want to just address it. Action really refers
`broadly to any act that a system executes based on the
`identification of a work. So, once the work is identified after the
`search, there is an action.
`What is -- what does Ghias tell us about that? If we
`look at slide 19, Ghias says, "the set of possible actions is
`potentially infinite," and we also asked Dr. Karypis in his
`deposition, and if we look at slide 20, we will see what he said,
`and we asked him the question, and he said, "I cannot really think
`of anything. I mean outside, you know, the constraints of the rest
`of the words in this step." That's not in the scope of what action
`actually means.
`So, I can't believe that there is going actually be a
`serious challenge on this, but if we look at the Ghias patent itself,
`on slide 21, it tells us that the database returns a list of songs, in
`the results of the query, the user can identify the song of interest.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`
` 12
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`So, that is the action, and I don't think there is going to be any
`really serious challenge to that.
`JUDGE PETTIGREW: Now, that takes place, at least
`the way it's presented there at the bottom of column 6 and top of
`column 7, after that, starting at line 5 of column 7, it says, "if the
`list is too large, the user can perform a new query on a restricted
`search list," which is one of the things that you said is the
`nonexhaustive search.
`MR. ELACQUA: That's correct.
`JUDGE PETTIGREW: I assume that you're implying
`that after that second search, there is also a display or an output of
`a list of songs.
`MR. ELACQUA: Yes, there could be an action as a
`result of each of the searches or at the end of the search.
`JUDGE PETTIGREW: All right, thank you.
`MR. ELACQUA: I would like to now move to the
`Iwamura reference, which is the second of our references. And
`here we have three issues, the nonexhaustive search issue,
`identifying a neighbor and sublinear. Now, there's a mistake on
`the slide here on slide 22, it should say three claims, not two
`claims for the '237 patent. And I'm going to preview Iwamura for
`just a moment.
`Iwamura represents musical works as a series of
`numbers, and those numbers are relative pitch differences
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 13
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`between adjacent notes. And to compare references, Iwamura
`computes the absolute differences between the query and the
`melody segments.
`Now, what's very important in understanding Iwamura
`is that Iwamura does not compare the query to every reference
`segment, but only makes comparisons where the peak notes
`alone, and peak note is defined as a note that is higher than or
`equal to each of the adjacent notes. So, we're dealing with a peak
`note algorithm.
`Now, the Patent Owner's position here is that matches
`equals songs and matches do not equal melody segments, but this
`is -- this is incorrect, for several reasons, and we'll see that if we
`look at the Patent Owner's response and the diagrams presented to
`illustrate certain positions by Dr. Karypis on slide 26, we will see
`that this is not correct. Because Dr. Karypis actually illustrates
`the nonexhaustive nature of Iwamura's search algorithm by
`putting segments and comparing the segments and actually taking
`the absolute differences which he totals up and circles. And then
`he shifts to the next peak and does the same thing.
`And, further, Iwamura nowhere in its disclosure,
`nowhere, calculates the overall difference or average difference
`between a query and a work over the entire length of a song. It's
`only done the way this is illustrated here, which is, again, from
`the Patent Owner's own response.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 14
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`Finally, at his deposition, Dr. Karypis confirmed that a
`single reference song may contain multiple matches. So, at his
`deposition, we put before Dr. Karypis a query, 5, 4, 3, 2, 1, on
`slide 27, and a reference. And he stated that in that reference, he
`found four portions that match exactly to the query. Now, the
`whole song cannot be the match. If he recognizes and considered
`that what he was matching to were segments and that he found
`four of them matched in that single reference point.
`So, his position on this is very consistent with
`Petitioner's position that we are matching melodies, we're not
`matching entire songs.
`JUDGE TURNER: Counselor, can I ask a quick
`question before you move on?
`MR. ELACQUA: Sure.
`JUDGE TURNER: In Iwamura, the database is made
`up of records; what are those records in Iwamura?
`MR. ELACQUA: The records in Iwamura is a string of
`numbers that represent the differences between the pitches in the
`notes of the song. And the only -- the only segments that are
`actually matched up against the query is when they align with the
`peak note in that string of numbers.
`JUDGE TURNER: But each record is a song. Is that
`correct?
`MR. ELACQUA: That's my understanding.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 15
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`JUDGE TURNER: So, and I guess the follow-up
`question I would have would be in Iwamura, isn't each record
`examined when you search?
`MR. ELACQUA: Each record -- each record or each
`song, yes, is looked at. The question is, in the search, is the
`match of the query to the song or is the match of the query to
`each segment that's aligned with the peak, and in Iwamura, it does
`not ever match the query, which can be, for example, a
`hummed -- a shortened hummed version of the song, or a, you
`know, a version that is played out in a short sequence on a piano
`or something. So, that short query is then compared only to the
`peaks in the notes. It's never compared to the entirety of the song.
`JUDGE TURNER: No, I understand that, but I guess
`my question would be, how is the basic search in Iwamura
`nonexhaustive if it looks at every single song? If it looks at every
`single record, how is that nonexhaustive?
`MR. ELACQUA: Because it's our position that when
`you're talking about possible matches, here in Iwamura, just like
`in Ghias, you're not looking to match the query to the entire song,
`because a match has to be the query to something. And the fact
`that they are just touching a note in the song wouldn't be
`sufficient to say that's actually how the match is being done,
`because the extracted features here are the extracted features are
`coming from the differences in the pitches of the song notes
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 16
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`themselves against each other. And what's being matched is the
`query, which is in this case, that we're showing on 27 an example
`of, would be, you know, five notes with the relative pitch
`differences, and it would only be lined up against the references
`that have the peak notes.
`JUDGE TURNER: So, would it be fair to say that
`Iwamura, the search is exhaustive with respect to the records, but
`nonexhaustive with respect to the features?
`MR. ELACQUA: It's nonexhaustive with respect to
`what is actually being matched. It's nonexhaustive with respect to
`what you're matching it to. You're not matching that query to the
`entire song at one time. You're going through a series -- a
`process which is a searching process of taking the query against
`each segment. That's what's being matched. Those are the
`possible matches that we're dealing with. We're not dealing with
`the entire song.
`We concede that every song is looked at.
`JUDGE TURNER: Okay.
`MR. ELACQUA: But we're not conceding that that's
`the match. That's not the match that we're talking about.
`JUDGE TURNER: Right.
`MR. ELACQUA: In fact, in a brute search, and in fact,
`we can show this. Andrew, can you put up the chart that we
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`
` 17
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`have, I will just show as an example, the square box chart. I
`apologize, we've lost the slide presentation.
`I think this might be helpful at this point. Again we're
`dealing with Iwamura and this is another instance where we
`presented this chart to Dr. Karypis at his deposition.
`JUDGE PETTIGREW: I'm sorry to interrupt, this is
`slide 65.
`MR. ELACQUA: Sixty-five. I'm sorry, Judge Turner,
`it's slide 65.
`JUDGE TURNER: Fifty-five?
`MR. ELACQUA: Sixty-five, thank you.
`JUDGE TURNER: Sixty-five, thank you. Got it.
`Thank you.
`MR. ELACQUA: So, here is the query, 5, 4, 3, 2, 1,
`and you see the reference base below it at the top of the box. If
`we were to do a brute force search or an exhaustive search, we
`would do each one of these 16 comparisons. But, in fact, because
`this is a peak note search, there's only two comparisons that are
`actually done, and the two comparisons are number 5, where you
`see the asterisk, because the 5 is lined up with the peak note 6 in
`the reference, which is asterisked at the bottom. And you see
`comparison 11, which is asterisk 5 and asterisk 6.
`And Dr. Karypis, as we will see, has admitted that those
`are the only two comparisons done and the other comparisons are
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 18
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`not complete. They're not -- they're never performed. And, in
`fact, you will also see, if you look at comparison 12, that's an
`exact match. We know that because if you look at the right-hand
`column, where the total absolute differences are, it lists zero. So,
`that shows that it's an exact match. That exact match is never
`performed and never found.
`So, if we were to do an exhaustive search using that
`query against the reference, that's what it would look like, 16
`comparisons and, in fact, the way Iwamura does it, there's only
`two.
`
`I will now move on to the neighbor search. Iwamura
`also has neighbor search as an issue. If we go back to slide 28,
`we will see that, again, we're using the same construction, of
`course, that we used with Ghias. And here, there can be no
`dispute that Iwamura locates a close match. We go to slide 29,
`we'll see that in Iwamura itself, it tells us that "the reference
`melody that gives the least difference is returned as a search
`result."
`
`The last reference that I want to discuss is Conwell, it's
`the last of the three references. And with Conwell, there's only
`two issues, identifying a neighbor, and nonexhaustive search.
`But to really understand Conwell, I believe it's important to
`understand hashing. And I want to make sure that we're all on the
`same page there.
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 19
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`So, I would like to provide just a quick review of that.
`Now, a hash is a mathematical algorithm that converts a large file
`into a much smaller number in a deterministic manner. The same
`file will always yield the same hash value. But there are other
`types of hash algorithms. For example, if we have slight
`differences in the input file, in some hash algorithms, you will
`still yield the same hash value. In others, a slight difference in
`the input value will yield a dramatically different hash value.
`So, when looking at Conwell, we first need to decide
`what type of hashing is used in Conwell. Conwell tells us, on
`slide 31, "of course, by suitably designing the algorithm by which
`identifiers are derived, nonidentical versions of the same basic
`content may nonetheless correspond to the same identifier."
`There is extensive published research on such technology, e.g.
`hashing algorithms by which similar or related, but non-identical,
`inputs map to the same hash outputs."
`So, we know that's the type of hashing that's being
`done, and the Board understood that and agreed in its institution
`decision for the 343 IPR at paper 6, page 11, and actually, the
`Patent Owner also agreed in the Patent Owner response, paper 17
`at 21.
`
`JUDGE PETTIGREW: Mr. Elacqua, you are into your
`rebuttal time, would you like to take a few more minutes and
`continue?
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 20
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Patent
`8,010,988 B2); Case IPR2015-00348 (Patent 8,656,441 B1)
`
`
`MR. ELACQUA: Yes, I would like to take a few more
`minutes and finish up here. I have just a few more things to say
`about Conwell. And one is that Conwell actually extracts a hash
`value from a query work and transmits it and then looks it up in
`identifier database using a standard database query. And based
`on that, there is a URL that is sent to the user.
`So, the question is, does Conwell, does it teach a
`neighbor search, and yes, it does teach a neighbor search, because
`the same hash can represent works that are similar, but not
`exactly the same, and this search can locate near matches. So,
`Conwell does perform a neighbor search. And if we look at slide
`34 quickly, the Board actually agreed with that.
`The final issue on Conwell that I will touch on is
`nonexhaustive, we're using the same construction. The sole
`dispute in Conwell is not whether or not the searches for matches
`are using a database query, but whether that database query is
`nonexhaustive. And the position that Patent Owner has been
`taking is that they argue that since it can also possibly be an
`exhaustive database query and not just a nonexhaustive one, it's
`not inherent, but inherency is not really the issue we're saying
`here.
`
`We're saying Conwell literally discloses a
`nonexhaustive search. And we know that because Dr. Cox, the
`inventor of these four patents, tells us. And if we look at slide 36,
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`20
`21
`22
`23
`24
`
` 21
`
`
`
`
`
`
`
`Case IPR2015-00343 (Patent 8,640,179 B1); Case IPR2015-
`00345 (Patent 8,205,237 B2); Case IPR2015-00347 (Paten