[00:00:03] Speaker 03: We have four argued cases this morning. [00:00:05] Speaker 03: The first of these is number 16, 2509, Google LLC versus Network One Technologies, Inc., Mr. Bagatell. [00:00:18] Speaker 00: May it please the court, Dan Bagatell, on behalf of Google, with me as my colleague Robert Swanson. [00:00:23] Speaker 00: I'd like to start, if I may, with the construction of non-exhaustive search, because if you agree with us on that, it will simplify the rest of the appeal considerably. [00:00:32] Speaker 00: This is a case in which the BRI standard makes a difference. [00:00:35] Speaker 00: It's a case where there are multiple reasonable interpretations and not much to go on beyond the claim language. [00:00:40] Speaker 00: The term non-exhaustive search doesn't appear in the specification. [00:00:43] Speaker 00: It was added years into prosecution with no explanation of the definition. [00:00:48] Speaker 00: There's no ordinary meaning, single ordinary meaning. [00:00:51] Speaker 02: Just to be clear, the term exhaustive also does not appear in the specification. [00:00:56] Speaker 00: It doesn't. [00:00:57] Speaker 00: The patent does refer to the term linear search, but it doesn't clearly answer [00:01:02] Speaker 00: the relevant question for purposes of this appeal. [00:01:06] Speaker 00: So we think the right question is the BRI standard. [00:01:08] Speaker 00: Is there a reason to rule out our definition, either intrinsic or extrinsic evidence? [00:01:13] Speaker 00: I'll start with the claim language. [00:01:17] Speaker 00: You mentioned exhaustive search. [00:01:18] Speaker 00: Exhaustive means to go through completely, to use up entirely. [00:01:23] Speaker 00: In the case of this type of a search, [00:01:25] Speaker 00: It means that you're considering every single one and making sure you're going to get every possible match. [00:01:30] Speaker 00: And in this case, there could be multiple matches because we're not talking about just a search that gives you one right answer. [00:01:36] Speaker 00: This is a neighbor search. [00:01:36] Speaker 00: You can get multiple answers, and they may not be exact matches. [00:01:40] Speaker 00: So you can't just judge it by the first character. [00:01:43] Speaker 00: We used the example of Quarter Heels for the Federal Circuit in the database, and you search on Federal Circuit. [00:01:48] Speaker 00: If you're looking for an F to match a C, it's not going to match. [00:01:51] Speaker 00: If you're looking for the first word, it's not going to match. [00:01:54] Speaker 00: characters may well be spurious. [00:01:56] Speaker 00: To get the right answer, you need to consider the whole thing. [00:01:59] Speaker 00: You have to consider it holistically. [00:02:00] Speaker 00: At least that's a reasonable interpretation of the claims. [00:02:03] Speaker 00: And there's nothing in the intrinsic or extrinsic evidence that suggests otherwise. [00:02:07] Speaker 02: I mentioned the... So let's say that you come up with a search and the Court of Appeals Federal Circuit comes up. [00:02:17] Speaker 02: Under your construction, does it stop there? [00:02:19] Speaker 02: It seems to me under your construction, [00:02:21] Speaker 02: You would have to go through and view all the data in there and make sure that Judge Raina is listed with all the judges and make sure the address is correct and that all the data within that particular result is correct or matches. [00:02:39] Speaker 00: Well, that was our construction that you need to look at them. [00:02:42] Speaker 00: It may or may not match depending on whether it is a near neighbor or a neighbor search. [00:02:46] Speaker 00: So for example, [00:02:48] Speaker 00: Federal Circuit will not exactly match Court of Appeals for the Federal Circuit, but it can be a near match. [00:02:53] Speaker 00: And that's our point, that you can't just stop looking at one bit of data or one note. [00:02:58] Speaker 00: You have to consider more than that. [00:03:01] Speaker 00: And there really isn't any reason in the specification to go otherwise. [00:03:04] Speaker 00: In terms of extrinsic evidence, the best the other side came up with was a Wikipedia definition of brute force search, which we'll call that equivalent to an exhaustive search. [00:03:13] Speaker 00: But what it says is check to see whether each candidate's solution satisfies the search criteria. [00:03:19] Speaker 00: Well, if anything, that helps us, because in the case of a search criteria where you're dealing with a neighbor search, you have to see whether it's going to match holistically, not just whether it's going to match literally character by the first character or the first bit. [00:03:32] Speaker 00: I recognize the board did not credit our expert, but you don't need to rely on an expert. [00:03:37] Speaker 00: If you had a situation where there's no clear answer from the interviewer. [00:03:41] Speaker 00: Neither expert here pointed to anything [00:03:44] Speaker 03: in the art that suggested what the definition was, right? [00:03:48] Speaker 03: They were just opining as to what they thought. [00:03:51] Speaker 00: That was true of your expert as well as theirs, right? [00:03:54] Speaker 00: Largely. [00:03:55] Speaker 00: I think the other side, well, first of all, the board decided this at the institution phase, so their expert hadn't even opined at that point. [00:04:04] Speaker 00: It was just lawyer argument, and in our case, an expert with a petition. [00:04:07] Speaker 00: And the board did not credit our expert. [00:04:09] Speaker 00: They basically just said we hadn't [00:04:12] Speaker 00: proven that you should include that particular element of the construction. [00:04:17] Speaker 00: It brought half of our construction, but not the other half. [00:04:20] Speaker 00: But the burden of proof wasn't on us. [00:04:22] Speaker 00: This was a legal issue governed by the BRI standard. [00:04:25] Speaker 00: And so if there are multiple reasonable interpretations, this is a difficult case in the sense that usually you have something in the intrinsic evidence that points you one way or the other or there's a dictionary definition. [00:04:35] Speaker 00: If you don't have that, then you have to take the broadest interpretation. [00:04:38] Speaker 00: I think the board got a little messed up because... I want to come back to my question. [00:04:42] Speaker 03: Did either, apart from the Wikipedia page, did either expert point to any documentation from the prior art suggesting what meaning this term should have? [00:04:56] Speaker 00: I don't recall any. [00:04:58] Speaker 00: And so that's why we're in a situation where [00:05:01] Speaker 00: you pretty much are going on the claim language. [00:05:03] Speaker 00: And if nothing rules out our reasonable interpretation, and again, you're dealing with a non-exhaustive search for a neighbor, it's not an exact search. [00:05:12] Speaker 00: You can't just stop it the first bit. [00:05:14] Speaker 00: At least we had a reasonable interpretation. [00:05:16] Speaker 00: And if you agree with us, that's going to require reversal of the claim construction and remand for reconsideration under the correct claim construction. [00:05:23] Speaker 00: I believe that affects both [00:05:25] Speaker 00: Iwamura and Gias. [00:05:27] Speaker 00: I'm not even sure there'll be a dispute as to whether they disclosed non-exhaustive search under our interpretation. [00:05:32] Speaker 00: I acknowledge though that Conwell really went off at an orthogonal ground and if you would like I can turn to Conwell or I can go to Iwamura under the board zone construction, whichever you're on. [00:05:43] Speaker 02: Before you do that, let me ask you another question. [00:05:48] Speaker 02: The 441 pan for example, it relates to identifying a work. [00:05:52] Speaker 02: such as a digital audio or video file without needing to modify that work. [00:05:58] Speaker 02: So it's looking for a work, a video file or a digital file, let's say music, right? [00:06:07] Speaker 02: It seems to me that there you're talking about a return that's not necessarily the result of a search for data. [00:06:21] Speaker 02: And your construction seems to be focused more on data as opposed to a record. [00:06:28] Speaker 02: And it seems to me if you look at the patent at appendix 99, column 8, and starting with that whole language 45 running into the next page, I think it's clear there that the patent is not speaking about data. [00:06:46] Speaker 02: In fact, it says that that's not what the patent is about. [00:06:50] Speaker 02: And it refers to things such as entries and walks away from having to make a binary search, just looking for zeros and ones as opposed to a note, a sound. [00:07:05] Speaker 00: Well, to some extent, it's dealing with some of the problems with existing ways to do searches. [00:07:12] Speaker 00: So some of this is a matter of semantics, whether you refer to it as data or records in a database. [00:07:17] Speaker 00: I've been trying to avoid some of the semantics and basically refer to, you've got to search on non-exhaustive features of a media work. [00:07:25] Speaker 00: That's the claim language for all of these. [00:07:27] Speaker 00: And so you're taking the media work, you're extracting certain features, which can be certain segments or it can be a mathematical manipulation. [00:07:37] Speaker 00: And then you're trying to see whether that matches something that the user inputs. [00:07:41] Speaker 00: Maybe they hum it, maybe they type it in, whatever. [00:07:44] Speaker 00: So in that sense, the other side has been focusing on the term database. [00:07:49] Speaker 00: I agree. [00:07:49] Speaker 00: It's not really a question of database. [00:07:51] Speaker 00: Database isn't the claim term. [00:07:53] Speaker 00: The question is whether you're extracting features from this media work and you're trying to find a near match to it or a neighbor search, non-exhausted neighbor search for a near match of that item. [00:08:04] Speaker 00: So we're not necessarily searching on the media works themselves. [00:08:08] Speaker 00: We're actually doing a search of the extracted features. [00:08:10] Speaker 00: And Iwamura is an example of that. [00:08:13] Speaker 00: you have this media work and maybe it's got the first 100 notes of a passage of Beethoven or something like that. [00:08:23] Speaker 00: And you enter a 10 note search. [00:08:26] Speaker 00: It's not going to be looking at all 100 notes. [00:08:29] Speaker 00: It's going to be looking at the manipulation of one through 10, two through 11, three through 12. [00:08:36] Speaker 00: It's a frame. [00:08:37] Speaker 00: It's going to be a moving frame. [00:08:38] Speaker 00: It's going to be looking at all the 10 note extracted features. [00:08:41] Speaker 00: And in fact, in this case, [00:08:43] Speaker 00: They're only going to be looking at roughly 20% of them, because it's looking for the ones that have aligned peaks. [00:08:49] Speaker 00: So 1 through 10, the peaks don't align. [00:08:51] Speaker 00: We're going to throw that out. [00:08:52] Speaker 00: 2 through 11, doesn't align. [00:08:53] Speaker 00: We're going to throw that out. [00:08:54] Speaker 00: 3 through 12, ah, the peaks align. [00:08:57] Speaker 00: Therefore, we're going to search that one. [00:08:59] Speaker 00: Even more, it teaches you that you can throw out roughly 80%. [00:09:02] Speaker 00: That is a non-exhaustive search, even under the board's definition, because the extracted features there are going to be the 10-note subsegments. [00:09:12] Speaker 00: You know, it may ultimately identify the Beethoven work or it may identify the hundred note sequence, but what it's actually searching is those ten notes. [00:09:23] Speaker 02: That somehow doesn't sound consistent with your argument as you're seeking to add an additional limitation to the court, to the board's construction. [00:09:33] Speaker 02: You're looking for all data within the possible matches. [00:09:36] Speaker 00: Yes. [00:09:36] Speaker 00: Well, okay. [00:09:37] Speaker 00: That's our claim construction. [00:09:39] Speaker 00: The argument I was just making was under the board's claim construction. [00:09:41] Speaker 00: Under our claim construction, we're basically trying to say when you're looking for a search of extracted features, if the extracted features are, for example, in these references, notes of a work, let's not call them data. [00:09:54] Speaker 00: Let's just call them notes or they could be adjusted notes. [00:09:59] Speaker 00: They could be relative pitches and that sort of thing. [00:10:02] Speaker 00: So they're adjustments of note data. [00:10:05] Speaker 00: Well, they're adjustments of notes. [00:10:06] Speaker 00: So you're going to see, does the first note match? [00:10:09] Speaker 00: Is that enough? [00:10:10] Speaker 00: Under their construction, that's enough. [00:10:11] Speaker 00: You can just say the first note doesn't match and throw it out. [00:10:14] Speaker 00: Under our construction, you need to consider it holistically. [00:10:17] Speaker 00: You take the extracted feature as a whole and compare it to each element, and that's a possible match. [00:10:24] Speaker 00: And in our view that you can't just stop at note one, letter one. [00:10:27] Speaker 00: The Federal Circuit is easiest to understand. [00:10:29] Speaker 00: I'm a terrible singer. [00:10:30] Speaker 00: You wouldn't want to hear me sing. [00:10:32] Speaker 00: So I think if you were to stop at the F, you would get a spurious result. [00:10:37] Speaker 00: And that's why you want to look holistically. [00:10:39] Speaker 00: So I think I've explained Ewa Murrah, I hope, to the Court's satisfaction. [00:10:44] Speaker 00: If you like, I can briefly touch on Conwell. [00:10:46] Speaker 00: The argument there is relatively straightforward. [00:10:48] Speaker 03: Conwell was a lookup in a large, sordid... Does Conwell get you anywhere under the board's construction? [00:10:59] Speaker 00: I don't think the board's construction is... The claim construction disagreement isn't really relevant to the way the board ruled on Conwell. [00:11:06] Speaker 00: I think [00:11:07] Speaker 00: Even if I win on the claim construction, I still need to convince you that the board erred with respect to the analysis of Conwell. [00:11:13] Speaker 03: But if you lose on claim construction, Conwell doesn't help you much, right? [00:11:17] Speaker 00: No, no. [00:11:18] Speaker 00: We have an argument even under the board's construction for Conwell, and I'll just give it to you in a nutshell. [00:11:24] Speaker 00: Conwell was a lookup in a large, sorted database. [00:11:29] Speaker 00: The person of ordinary skill in the art here had a comp sci degree and three years of experience. [00:11:34] Speaker 00: You sort a large database so that you can search it more efficiently. [00:11:37] Speaker 00: Their expert admitted as much. [00:11:40] Speaker 00: It didn't use the term non-exhaustive search, neither did this patent. [00:11:43] Speaker 00: And perhaps it was not an inherent anticipation because theoretically you could plod through a million records one by one. [00:11:52] Speaker 00: But a skilled artisan would know that you searched a large sorted database [00:11:56] Speaker 00: by using a non-exhaustive search. [00:11:58] Speaker 00: That's what anyone would do. [00:11:59] Speaker 00: We were relying on the doctrine that basically you immediately envisage it. [00:12:04] Speaker 00: It's not an abscissimus verbus test. [00:12:06] Speaker 00: It's a question of what one's skill in the art would understand looking at the reference. [00:12:09] Speaker 00: And the only thing you would do with a large sort of database is sort it non-exhaustively. [00:12:14] Speaker 00: I see I've exhausted my initial time. [00:12:16] Speaker 00: I'll save the rest for rebuttal if I may. [00:12:18] Speaker 03: OK. [00:12:18] Speaker 03: Thank you, Mr. Baggett. [00:12:23] Speaker 03: Is that how you pronounce it? [00:12:24] Speaker 01: I'm sorry? [00:12:25] Speaker 03: Devel, is that how you pronounce it? [00:12:27] Speaker 01: Devel, yes, your honor. [00:12:30] Speaker 01: Under this court's holding in Microsoft v. Polycon, a construction is not reasonable if it is divorced from the specification or if it is inconsistent with how one of ordinary skill would use a term. [00:12:47] Speaker 03: Let's talk about how one of ordinary skill would use the term. [00:12:51] Speaker 03: I've read the [00:12:52] Speaker 03: expert testimony here, and I don't see that there's anything very meaningful in that expert testimony. [00:13:00] Speaker 03: It says, well, this is the way I would construe it, but it doesn't really provide any evidence that this is a term that is understood in the art, right? [00:13:11] Speaker 01: I disagree, Your Honor, and the reason is our expert's testimony goes through the specification to explain in this context of the specification how it's being used. [00:13:22] Speaker 01: And if I might, context is crucial here. [00:13:25] Speaker 03: But hold on. [00:13:26] Speaker 03: Just put aside the specification for a moment. [00:13:29] Speaker 03: Is there any evidence that was presented that tells us that this has a meaning in the art? [00:13:35] Speaker 01: Yes, Your Honor. [00:13:36] Speaker 01: Our expert opined that it had a meaning in the art, and he also provided an example of an objective source, the Wikipedia definition, that the board accepted as an objective source indicating the ordinary meaning of exhaustive. [00:13:52] Speaker 03: The Wikipedia entry is the only piece of evidence that he relied on, putting aside the specification, as to what the meaning of this was. [00:14:02] Speaker 01: Directly. [00:14:03] Speaker 01: He did indirectly rely upon other articles to support his conclusion about the non-exhaustive searches in the specification. [00:14:11] Speaker 03: So what is it in the Wikipedia article that's so helpful to you? [00:14:15] Speaker 01: Your Honor, I'm going to turn to the Wikipedia article, but let me give you [00:14:20] Speaker 01: an explanation of what we're looking for in that article, one way or the other. [00:14:23] Speaker 01: The term non-exhaustive or exhaustive, it depends strictly on context. [00:14:27] Speaker 01: For example, if you're trying to locate where did you park your car in a parking structure, you do an exhaustive search if you check every parking space. [00:14:36] Speaker 03: Well, let's talk about the Wikipedia article. [00:14:38] Speaker 03: Where do we find the Wikipedia article? [00:14:40] Speaker 01: Appendix 1393. [00:14:41] Speaker 01: That's in the volume one of the joint appendix. [00:14:45] Speaker 01: 1393. [00:14:46] Speaker 01: 1393. [00:14:46] Speaker 01: OK, so how does this help? [00:14:50] Speaker 01: And we'll see that what it says in the first paragraph, in computer science, exhaustive search consists of systematically enumerating all possible candidates for the solution and checking whether each candidate satisfies the problem statement. [00:15:09] Speaker 01: What that means is we will have exhausted if we determine whether each candidate satisfies [00:15:17] Speaker 01: our algorithm. [00:15:17] Speaker 01: Why isn't that equally consistent with looking at all the data? [00:15:21] Speaker 01: Because, Your Honor, we do not need to look at all the data to determine whether each candidate satisfies the problem. [00:15:28] Speaker 01: We need to look at enough data to know whether it matches our algorithm. [00:15:32] Speaker 01: The example I gave Your Honors of looking for your parked car, you don't have to walk around every car. [00:15:37] Speaker 03: It may be or may not be, but I don't see how this is helping you. [00:15:42] Speaker 03: Because it doesn't really say [00:15:45] Speaker 03: that where it's talking about all candidates that doesn't mean looking at the data for all the candidates. [00:15:51] Speaker 01: It necessarily does, Your Honor. [00:15:52] Speaker 01: And the reason is this. [00:15:53] Speaker 01: If the question is this, are we trying to figure out whether a given candidate matches or not, we need to look at enough data to make that determination. [00:16:03] Speaker 01: That's always the case with every type of exhaustive search. [00:16:07] Speaker 01: And as Judge Raina pointed out, in the context of this patent, that's why the specification decides this issue. [00:16:13] Speaker 01: We can't divorce it from the specification. [00:16:15] Speaker 01: Our context is we're looking at a database to determine which entries or records match. [00:16:23] Speaker 01: We will have done that if we've examined every entry or record in the same way that we will have examined every parking space, if we look at every parking space. [00:16:33] Speaker 01: We don't have to walk around every car, look at it from the front, the back, and the sides to do an exhaustive search for a parked car. [00:16:42] Speaker 01: We don't have to look at every bit of data in a record to do an exhaustive search to determine whether each record matches or not. [00:16:52] Speaker 01: That's how exhaustive is used in the art. [00:16:54] Speaker 01: That's the ordinary meaning of it. [00:16:56] Speaker 01: And the proof of that, Your Honor... We don't have any evidence of that. [00:16:58] Speaker 01: We do, Your Honor. [00:16:59] Speaker 01: We have this. [00:17:01] Speaker 01: We have how it's used in the specification. [00:17:03] Speaker 01: And we have a complete app. [00:17:04] Speaker 03: What is it in the specification that tells you that this Google construction is incorrect? [00:17:10] Speaker 01: Your Honor, again, it's about the context. [00:17:14] Speaker 01: The context of every embodiment in the specification, I'll walk your Honors through it, is about looking for an entry in a database. [00:17:22] Speaker 01: So if we start with column 6, this is appendix page 98. [00:17:27] Speaker 01: We use this from the 179 patent as our specification. [00:17:32] Speaker 01: We start at appendix page 98, column 6. [00:17:37] Speaker 01: We go down to line 21. [00:17:39] Speaker 01: What we have is this work identification information storage 110. [00:17:43] Speaker 01: That's a database, it explains. [00:17:46] Speaker 01: It includes a number of items or records, 112, each of which we associate a feature vector. [00:17:53] Speaker 01: The vector is the group of data we're going to be looking at. [00:17:56] Speaker 01: So the whole context here, and we're going to walk through it, is always about we've got a database, [00:18:01] Speaker 01: It's organized by records, each of which has a set of data, a vector that we're looking at. [00:18:06] Speaker 01: Let's go over to column 8 then, the very next page, starting at line 49. [00:18:10] Speaker 01: Here we're going to have something we're trying to compare, and we're going to compare it to entries of known vectors 114 in a content identification, WID database 110. [00:18:22] Speaker 01: The context is trying to find out which entries in the database match. [00:18:28] Speaker 01: Turn over to the next page, column 9, starting at line 38. [00:18:32] Speaker 01: If the extracted vector matches a known vector in the content identification base, then the work has been identified. [00:18:42] Speaker 01: So what we're trying to determine is which entries in this database match. [00:18:48] Speaker 03: And in that context- It may be that the specification is describing non-exhaustive searches the way you're describing them. [00:18:57] Speaker 03: But that doesn't mean that an exhaustive search doesn't look at all the data. [00:19:05] Speaker 03: And that other things might not qualify as non-exhaustive searches that don't look at all the data. [00:19:13] Speaker 01: It does by definition, Your Honor. [00:19:14] Speaker 01: Again, exhaustive always depends on context. [00:19:18] Speaker 01: If you're going to do an exhaustive search for a car, that's different than an exhaustive search for your car keys. [00:19:23] Speaker 01: It requires looking at different data. [00:19:24] Speaker 03: But this doesn't say that the only kind of non-exhaustive searches are the searches that are described in this part of the specification. [00:19:31] Speaker 01: No, but what it does tell us is the context. [00:19:34] Speaker 01: What are we trying to find? [00:19:37] Speaker 01: If we're trying to find items in a database that match, we are exhausting if we've looked at every item. [00:19:44] Speaker 01: We don't have to look at every bit of data in an item to determine whether or not it matches. [00:19:48] Speaker 01: We have to look at enough data to satisfy or dissatisfy our algorithm. [00:19:55] Speaker 01: Our expert opined that way. [00:19:57] Speaker 01: We've got an objective source, and importantly, your honor, [00:20:00] Speaker 01: Their construction is completely divorced from the intrinsic record. [00:20:06] Speaker 01: Under this court's controlling authority, Microsofty, Proxycon, and the subsequent cases, if it's divorced from the specification, it can't be reasonable. [00:20:15] Speaker 01: They don't point to a single line. [00:20:17] Speaker 03: If it's divorced from a specification that doesn't tell you what the term means, it's not reasonable? [00:20:22] Speaker 01: No, Your Honor. [00:20:23] Speaker 01: It's not that it doesn't have to have a definition in the specification. [00:20:26] Speaker 01: It's got to be in this context. [00:20:28] Speaker 01: In this context, [00:20:29] Speaker 01: There is no suggestion that to search exhaustively, we must look at all data. [00:20:35] Speaker 01: It specifically identifies as an exhaustive search, a linear search of all n entries. [00:20:42] Speaker 01: If we've examined each entry, we've exhausted. [00:20:46] Speaker 01: We don't have to look at each bit of data to exhaust. [00:20:50] Speaker 01: We have to examine each entry. [00:20:52] Speaker 01: The converse is also shown here. [00:20:55] Speaker 01: The very next paragraph on column 9, line 413, it identifies non-exhaustive searches. [00:21:03] Speaker 01: Each one of those searches is one that has a mechanism for pruning out and not looking at records. [00:21:10] Speaker 01: So all the examples of non-exhaustive search, every single one of them, is an example where it's going to prune out certain records. [00:21:18] Speaker 01: Those two things, and the board made a factual finding that this paragraph here [00:21:23] Speaker 01: does explain exhaustive and non-exhaustive searches. [00:21:27] Speaker 01: That factual finding is supported by our expert's testimony. [00:21:30] Speaker 01: The court has to accept that factual finding. [00:21:33] Speaker 01: Therefore, these passages in the specification do describe exhaustive and non-exhaustive searches. [00:21:38] Speaker 01: In addition, the board made a factual finding as to the ordinary meaning. [00:21:43] Speaker 01: That Wikipedia article did express the ordinary meaning. [00:21:47] Speaker 01: That factual finding is supported [00:21:50] Speaker 01: by our expert testimony and by the Wikipedia article itself. [00:21:55] Speaker 03: Do you recall that the board relied on the expert testimony and reached its conclusion? [00:22:00] Speaker 01: Well, Your Honor, it reached an initial conclusion. [00:22:03] Speaker 01: Did it rely on expert testimony? [00:22:06] Speaker 01: Yes, Your Honor, I'll explain. [00:22:08] Speaker 01: Where? [00:22:09] Speaker 01: If I can explain, it reached its initial conclusion and then... Where did it rely on expert testimony? [00:22:14] Speaker 01: Tell me. [00:22:15] Speaker 01: If we look at [00:22:22] Speaker 01: Record page 48. [00:22:24] Speaker 01: 48? [00:22:27] Speaker 01: Yes. [00:22:28] Speaker 01: Appendix 48. [00:22:29] Speaker 01: This is the decision, an example of one of the decisions, and what it does here, it recites its, this is the top of the, first full paragraph at the top of the page, it recites its construction, then it says, for this final written decision, after considering the complete record, we maintain our construction of a non-exhaustive search. [00:22:50] Speaker 01: That's the reference to the expert testimony? [00:22:53] Speaker 01: Yes, your honor. [00:22:53] Speaker 01: But to determine whether there's substantial evidence to support the board's finding, we don't look for sites in the record. [00:23:01] Speaker 01: The board doesn't have to cite it. [00:23:03] Speaker 01: It just has to be in the record. [00:23:04] Speaker 01: That's this court's consistent holding in case after case. [00:23:08] Speaker 01: It's not a question of, gee, well, the board didn't cite the expert. [00:23:12] Speaker 01: Therefore, we can ignore it. [00:23:13] Speaker 01: No. [00:23:13] Speaker 01: The question is, the board made a factual finding. [00:23:16] Speaker 01: Is there evidence in the record that supports that factual finding? [00:23:20] Speaker 01: Our expert's testimony does support it. [00:23:24] Speaker 01: So there's factual findings about the ordinary meaning. [00:23:27] Speaker 01: There are statements in this specification. [00:23:29] Speaker 01: They have nothing in the record, nothing to support their construction. [00:23:35] Speaker 01: Without that, it's completely divorced from both the specification and the evidence. [00:23:40] Speaker 01: And therefore, under this court's consistent holdings, it cannot be a reasonable interpretation. [00:23:46] Speaker 01: And in fact, as I've explained, it would be unreasonable. [00:23:48] Speaker 01: It would be unreasonable to say, [00:23:49] Speaker 01: We're trying to determine which records match. [00:23:54] Speaker 01: And we're also going to require for this something to be exhaustive, that it has to look at every bit of data, even though it makes no sense to look at that data. [00:24:01] Speaker 01: It would be utter nonsense. [00:24:03] Speaker 01: In fact, under their construction, it's hard to conceive of any actual algorithm that's used that would be exhaustive. [00:24:12] Speaker 01: And that's not how the term is used in the art, which is why they were unable to, after our expert presented his testimony, [00:24:19] Speaker 01: come forward with any example, in any patent, in any article, in any publication, anywhere, using non-exhaustive or exhaustive as they would use it. [00:24:31] Speaker 01: Not one. [00:24:32] Speaker 01: It's completely divorced from the evidence. [00:24:36] Speaker 01: Your Honor, let me turn to Iwamura, the first reference that they argue about. [00:24:42] Speaker 01: The board made a factual finding as to what were the possible matches in Iwamura. [00:24:48] Speaker 01: The board made a factual finding that the possible matches were the extracted features corresponding with a reference melody. [00:24:55] Speaker 01: To explain this, it may be helpful to look at an example. [00:24:58] Speaker 01: In our brief on page 36, we have this example of the extracted features from Iwamura. [00:25:06] Speaker 01: In this example, there are 20 numbers that represent a vector or extracted features that's used in Iwamura. [00:25:15] Speaker 01: And the board made a factual finding that [00:25:18] Speaker 01: the attractive features that matter are the ones that correspond to a reference melody, these 20 numbers. [00:25:24] Speaker 01: Google asserts that what matters is that Ewan Murrow looks at a subset of those. [00:25:30] Speaker 01: And a subset, they say, is going to be a possible match. [00:25:34] Speaker 01: But the board made a factual finding that the possible matches were the entire reference melody. [00:25:39] Speaker 01: That factual finding is supported by substantial evidence, including a direct admission from their own expert. [00:25:46] Speaker 02: Counsel, let me ask you a question. [00:25:48] Speaker 02: A very fundamental, basic question here. [00:25:52] Speaker 02: Don't you agree that Google's construction of non-exhaustive searches is broader than your construction or the construction of the board? [00:25:59] Speaker 01: It is broader. [00:26:00] Speaker 01: Yes, Your Honor. [00:26:02] Speaker 01: It is broader. [00:26:03] Speaker 02: That's why there are... Then what's your problem with it? [00:26:05] Speaker 02: Why would it be unreasonable if we are looking for the broadest reasonable interpretation? [00:26:11] Speaker 01: There are three reasons, Your Honor. [00:26:13] Speaker 01: First of all, it's unreasonable because it's divorced from the specification. [00:26:17] Speaker 01: any reasonable construction. [00:26:18] Speaker 01: This is the court's consistent holdings. [00:26:20] Speaker 02: For example... But we've already decided that the specification supports no definition, either for non-exhaustive or exhaustive. [00:26:29] Speaker 01: Well, Your Honor, there's a difference here. [00:26:32] Speaker 02: It doesn't have a... Let's put the specification aside, but what else do you have? [00:26:36] Speaker 01: It also has to be consistent with how one of ordinary Skittle New York would use it. [00:26:41] Speaker 01: It can't be divorced from that. [00:26:42] Speaker 02: It can't be inconsistent. [00:26:43] Speaker 02: And you use a Wikipedia site as authority for that? [00:26:47] Speaker 01: And our expert's testimony, Your Honor. [00:26:50] Speaker 02: Both of those... But there's no showing that the Board relied on your expert testimony. [00:26:55] Speaker 01: That's irrelevant, Your Honor. [00:26:57] Speaker 01: Under this Court... We just said it was relevant. [00:26:59] Speaker 01: No. [00:26:59] Speaker 01: It's not relevant whether there's a site to the Board. [00:27:03] Speaker 01: What the Board says is that it considered the complete record... But as a reviewing Court, [00:27:08] Speaker 02: And we're reviewing the board's decision. [00:27:11] Speaker 02: I am looking for a language. [00:27:13] Speaker 02: I just can't guess or assume that the board took everything into consideration. [00:27:18] Speaker 01: Yes, Your Honor, you can. [00:27:23] Speaker 01: And I just cite, Your Honor, to Siemens Energy v. United States. [00:27:27] Speaker 01: It has to be supported by substantial evidence determined on the entirety of the record. [00:27:33] Speaker 01: Whether the board actually cites our expert or not. [00:27:35] Speaker 02: That's what I'm looking for. [00:27:36] Speaker 02: What else do you have? [00:27:39] Speaker 01: Well, we rely upon the specification, our expert's testimony, and the Wikipedia article. [00:27:45] Speaker 01: All of those things together support our construction. [00:27:48] Speaker 01: By contrast, there's not one page in this record they can support to, to suggest that exhausted in the context of a database search, reverse to look into all data. [00:28:00] Speaker 01: It is completely divorced from the specification. [00:28:03] Speaker 02: Well, other than the fact that you've conceded or have agreed with me, [00:28:07] Speaker 02: that the BRI here is Google's construction. [00:28:11] Speaker 01: We've not conceded that. [00:28:12] Speaker 01: It's not the broadest reasonable construction. [00:28:14] Speaker 01: It's broader. [00:28:15] Speaker 01: It is not reasonable. [00:28:17] Speaker 01: Again, under this court's holdings in Microsoft eProxyCon and man-machine interface, it has to be reasonable. [00:28:25] Speaker 01: And to be reasonable, it must not be divorced from the specification and the record evidence. [00:28:32] Speaker 01: That's a quote from those cases. [00:28:33] Speaker 03: It can't be that if the specification provides you no help, [00:28:37] Speaker 03: that it can't be that both parties' definitions are not supported one way or the other by the specification. [00:28:45] Speaker 01: No, that's not true. [00:28:46] Speaker 03: You can't look to ordinary meaning. [00:28:48] Speaker 01: Your Honor, there's an issue here. [00:28:51] Speaker 01: It's not that we need a definition in the specification. [00:28:54] Speaker 01: It has to be in the light of the specification. [00:28:56] Speaker 01: What is this patent trying to do? [00:28:58] Speaker 03: The specification doesn't even use the word exhaustive. [00:29:01] Speaker 03: It doesn't tell us what that means or non-exhaustive either. [00:29:05] Speaker 01: It does not, Your Honor, but we have [00:29:07] Speaker 01: factual finding by the board as two passages that describe exhaustive and non-exhaustive searches. [00:29:13] Speaker 01: That factual finding by the board is supported by substantial evidence. [00:29:17] Speaker 01: So the specification does identify exhaustive and non-exhaustive searches. [00:29:22] Speaker 01: Moreover, it gives us a context. [00:29:25] Speaker 03: Nobody's disputing that the specification describes some non-exhaustive searches. [00:29:30] Speaker 03: The problem is it doesn't tell you what the universe of non-exhaustive searches is. [00:29:34] Speaker 01: But it does give us a context, Your Honor. [00:29:37] Speaker 01: And context is crucial. [00:29:40] Speaker 01: This is not a specification where we're trying to do a non-exhaustive search of cars in a parking structure. [00:29:48] Speaker 01: We're trying to do a non-exhaustive search of items in a database. [00:29:53] Speaker 01: And in that context, the only reasonable construction is that we must check every entry in the database. [00:29:59] Speaker 01: We must check all items. [00:30:01] Speaker 02: By definition, if I do a statistical search, [00:30:05] Speaker 02: of records in a database, is that, by definition, non-exhaustive? [00:30:13] Speaker 01: No, Your Honor. [00:30:14] Speaker 01: It doesn't tell us one way or the other. [00:30:15] Speaker 01: You've given me a search algorithm. [00:30:17] Speaker 01: Now you need to tell me, are you going to look at every record in the database or not? [00:30:21] Speaker 01: That algorithm has a need. [00:30:22] Speaker 02: So the use of the term in the specification statistical search, that doesn't mean anything? [00:30:28] Speaker 01: No. [00:30:28] Speaker 01: It doesn't tell us whether that is exhaustive or non-exhaustive. [00:30:32] Speaker 01: You can have a statistical algorithm that's applied exhaustively or non-exhaustively. [00:30:38] Speaker 01: For example, their own expert, Dr. Moline, in his declaration at appendix 1132, he says, because neighbor searching is computationally intensive, content recognition schemes typically employed search algorithms that increased efficiency. [00:31:00] Speaker 01: by intelligently searching only a subset of potential matches, i.e., non-exhaustive algorithms. [00:31:08] Speaker 01: He equates non-exhaustive with intelligently searching only a subset of potential matches in the context of neighbor searching. [00:31:19] Speaker 01: So we could do a neighbor search, and it can be non-exhaustive if we have carved out certain potential records and not examined them at all. [00:31:28] Speaker 01: How do we do that? [00:31:30] Speaker 01: using techniques such as a KD tree and so on. [00:31:33] Speaker 03: Okay, I think Mr. DeValle, we're out of time. [00:31:35] Speaker 03: Thank you. [00:31:36] Speaker 03: Thank you, Your Honor. [00:31:40] Speaker 00: Mr. Bagatelli, you've got about three minutes. [00:31:43] Speaker 00: I'd like to just address a couple of points and then finish with a housekeeping matter, if I may. [00:31:48] Speaker 00: The first point I'd like to make is that the Board did not find an ordinary meaning of non-exhaustive search. [00:31:52] Speaker 00: It simply said, we're not convinced by your expert. [00:31:56] Speaker 00: The second point I'd like to make is that [00:31:58] Speaker 00: My colleague has pointed to a case of this court that says, basically, substantial evidence considers everything in the record, even if the board doesn't consider it. [00:32:06] Speaker 00: That's not an accurate portrayal of this court's case law. [00:32:09] Speaker 00: We've got a chennery doctrine. [00:32:10] Speaker 00: We rely on findings that the board actually made. [00:32:13] Speaker 00: The case on which he's relying is simply, it cites the Supreme Court's universal camera case, and it basically says, you look at everything in the record. [00:32:18] Speaker 00: In other words, what both sides have submitted. [00:32:22] Speaker 00: And in a jury verdict situation, you can [00:32:24] Speaker 00: infer maybe that the jury relied on something. [00:32:26] Speaker 00: In the case of an agency decision, you've got to have findings by the board. [00:32:30] Speaker 00: The board did not make findings about the ordinary meaning of non-exhaustive search. [00:32:37] Speaker 00: We do agree that it depends on context, but the context here is not a parking space. [00:32:43] Speaker 00: The context here is something, for example, where you're comparing two melodies. [00:32:46] Speaker 00: I'm a terrible singer. [00:32:47] Speaker 00: I enter things wrong. [00:32:48] Speaker 00: You've got to be able to figure out whether that's a near match. [00:32:51] Speaker 00: and you can't do that based on one. [00:32:52] Speaker 00: Their construction, the board's construction says one note, one bit is enough. [00:32:58] Speaker 00: My colleague is saying that, well, you don't necessarily need all of them. [00:33:01] Speaker 00: Well, that may show that there's a range of reasonable interpretations, but this is a BRI standard here. [00:33:06] Speaker 00: This is a case where the BRI standard really does make a difference. [00:33:13] Speaker 00: Iwamura, just one quick point on that, and that is that Iwamura, yes, it does talk about [00:33:20] Speaker 00: basically a 20 note segment that's going to be one of the reference melodies. [00:33:26] Speaker 00: But it's not comparing the whole 20 notes. [00:33:28] Speaker 00: It's comparing the 10 notes that you enter. [00:33:31] Speaker 00: And it says to disregard everything that's basically you're aligning two arrays and disregarding everything that doesn't line up. [00:33:36] Speaker 00: And then he's telling you, don't search all of those. [00:33:40] Speaker 00: Don't search 1 through 11 or 2 through 12, 3 through 13. [00:33:44] Speaker 00: Only do the ones where it lines up. [00:33:46] Speaker 00: So he's only searching 80%. [00:33:48] Speaker 00: So really, you have to do an apples-to-apples search, not an apples-to-orange search. [00:33:52] Speaker 00: You don't do 10 against 20. [00:33:53] Speaker 00: You're searching extracted features of 10. [00:33:57] Speaker 00: If I may, I'd just like to mention the housekeeping matter. [00:34:02] Speaker 00: There is a little bit of discrepancy in the briefs about which claims are on appeal and which references go with which claims. [00:34:10] Speaker 00: I've discussed that with opposing counsel. [00:34:12] Speaker 00: And what I would do is propose to submit a letter from the parties jointly. [00:34:16] Speaker 00: I don't think there's a real disagreement. [00:34:19] Speaker 00: clarify for the court which claims are at issue and which ones go with which references. [00:34:23] Speaker 00: Otherwise, it may be hard to parse out when drafting a decision. [00:34:28] Speaker 00: You can submit a joint letter. [00:34:30] Speaker 00: Thank you, Mr. Magatow. [00:34:31] Speaker 03: Thank both counsels.