[00:00:00] Speaker 01: The next case for argument is 23-1911 Xerox Corporation versus Meta Platforms. [00:00:09] Speaker 01: Mr. Powell, good morning. [00:00:10] Speaker 00: Good morning. [00:00:13] Speaker 00: Please proceed. [00:00:15] Speaker 00: Good morning, Your Honors. [00:00:16] Speaker 00: May it please the Court, Charles Fowler for Xerox Corporation. [00:00:20] Speaker 00: I plan to focus this morning on three issues. [00:00:24] Speaker 00: First, [00:00:26] Speaker 00: that the asserted prior art does not teach the previously unseen limitation in the 781 claim. [00:00:33] Speaker 00: And that's got two parts. [00:00:36] Speaker 00: One, Metta's primary reference, Koenig. [00:00:39] Speaker 00: And then second, I'll address the lack of a motivation to combine Koenig with Metta's secondary reference, Van Vliet. [00:00:49] Speaker 00: And then the third issue that I plan to address this morning is the lack of a motivation to combine [00:00:55] Speaker 00: Metta's primary reference with manbur, which was used to teach the separate threshold distance limitation in 781 claims. [00:01:06] Speaker 00: So beginning with the previously unseen element and the primary reference conic. [00:01:13] Speaker 00: Meadow's entire argument with respect to Koenig teaching a previously unseen element focuses on a short disclosure at column 24 of the Koenig reference, which is a patent that talks about applying Koenig's user model to predict interest in unseen documents. [00:01:36] Speaker 00: And the key error that Meta and the Broke made here is that unseen and- It doesn't quite say that. [00:01:44] Speaker 03: It says actually applying the user model to unseen documents. [00:01:51] Speaker 00: That's right, Your Honor. [00:01:52] Speaker 00: And what the user model does in this contest is it's talking about calculating [00:01:59] Speaker 00: P of U given D, which is an estimate of user interest in the document. [00:02:04] Speaker 00: So that's the application of the user model as applied to unseen documents. [00:02:11] Speaker 00: And the error here is that unseen in coding is not about a limitation or exclusion. [00:02:21] Speaker 00: It is about an ability to address documents that the system has not seen before. [00:02:28] Speaker 01: First, I guess I'm trying to understand what the point of your argument is. [00:02:33] Speaker 01: Word characterized it as saying that you contend that CONAC never expressly limits P or D to just unseen documents. [00:02:43] Speaker 01: Is that the essence of your argument, that it's not limited to just unseen documents? [00:02:48] Speaker 00: Correct, Your Honor. [00:02:49] Speaker 00: And the claim, however, does require limiting to only unseen documents. [00:02:56] Speaker 03: It's that only. [00:02:58] Speaker 03: That's the key thing you were talking about. [00:02:59] Speaker 00: That's the key. [00:03:00] Speaker 00: Correct, Your Honor. [00:03:01] Speaker 00: The key point in the claim, the key language in the claim, is only documents previously unseen by the user. [00:03:08] Speaker 00: CONIG does not do that. [00:03:10] Speaker 00: We don't dispute that. [00:03:11] Speaker 01: But isn't it sort of our basic obviousness law that you don't have to say only? [00:03:16] Speaker 01: I mean, if the prior art consists of [00:03:20] Speaker 01: the fact that it can be applied to unseen documents. [00:03:23] Speaker 01: Isn't that relevant prior art for the purpose of obviousness? [00:03:27] Speaker 01: Even if it has another section instead. [00:03:29] Speaker 01: And if you don't apply it, if you apply it otherwise, then you have another analysis. [00:03:35] Speaker 01: Is that not the way the law works? [00:03:37] Speaker 00: I don't think so, Your Honor. [00:03:39] Speaker 00: I don't think it would be obvious to a person of ordinary skill to do what our claim does, which is build a neighborhood to search in that excludes anything the user has seen before unless the person of ordinary skill is looking at a reference that does that. [00:03:59] Speaker 00: And what our point is with respect to Koenig is that [00:04:03] Speaker 00: One, we think Konig applies the user model to both seen and unseen. [00:04:11] Speaker 00: Two, Konig expressly teaches with respect to this personal crawler that Meta relies on not limiting the search universe to previously unseen documents, and the way it does that [00:04:26] Speaker 00: And this is at column 27. [00:04:28] Speaker 00: This is where the code of reference actually describes the search functionality that Meta relies on as teaching most of the challenge claim. [00:04:41] Speaker 00: It explains that the user model is applied to score each linked document. [00:04:52] Speaker 00: So what it's talking about there is [00:04:55] Speaker 00: Conan takes what it calls an initialized stack of documents that are things the user has almost certainly seen before. [00:05:03] Speaker 00: It's documents that are of high interest to the user, like browsing history, bookmarks, things like that, and analyzes the links in those documents and crawls for other documents. [00:05:14] Speaker 00: And it applies the user model to score each document. [00:05:21] Speaker 00: By using the word each there, it confirms that it is not excluding documents that the user has seen before from what it's applying. [00:05:34] Speaker 00: the PUD Formula 2, and that undermines Mehta's argument. [00:05:39] Speaker 00: Reinforcing that is the prosecution history, which we emphasized before the board, and the board did not address it all. [00:05:46] Speaker 00: The prosecution history confirms that Unseen and Koenig is about ability or capability, not limitation. [00:05:57] Speaker 00: There is a section, and this is at Appendix, I think it's 2448, [00:06:04] Speaker 00: 2448 and 49 where the inventors are distinguishing prior art search systems that are based on what it called a memorization model of [00:06:17] Speaker 00: In contrast, the invention in Koenig uses what I call a prediction or generalization model that unlike the prior art was able to deal with documents that the system had not seen before. [00:06:32] Speaker 00: And as the inventors explained, quote, there is no need to distinguish between seen and unseen documents. [00:06:40] Speaker 00: That's because [00:06:42] Speaker 00: The Koenig system can and does deal with them both equally and that was what separated Koenig from the prior art. [00:06:49] Speaker 00: What do you mean deal with them equally? [00:06:51] Speaker 00: What does that mean? [00:06:53] Speaker 00: Analyze them. [00:06:55] Speaker 00: In other words, so Koenig applies its user model to predict user interest, and it only includes in the stack of documents at its crawling step, documents that exceed a certain predicted level of user interest. [00:07:10] Speaker 00: So Koenig can do that whether it's seeing the document or not seeing the document, and that's the key difference between Koenig and the prior art. [00:07:18] Speaker 00: but was being distinguished when the inventors explained there's no need to distinguish between seen and unseen documents. [00:07:26] Speaker 03: And can you remind me that the board, did the board find the only unseen documents independently in Koenig and in Koenig combined with either, was it Mamber or Van Fleet? [00:07:43] Speaker 00: It did, Your Honor. [00:07:44] Speaker 00: The board used Van Vliet as a sort of alternative ground, and I'll turn to that one now. [00:07:50] Speaker 03: And your challenge to that has to do with the motivation to combine? [00:07:55] Speaker 00: Yes, Your Honor. [00:07:55] Speaker 00: The key challenge there is motivation to combine, and to frame that issue, [00:08:03] Speaker 00: Coding, similarly to the 781 patent, proceeds in two steps. [00:08:08] Speaker 00: It does what it calls a crawling to build a stack of documents to search, which Meta says is equivalent to our proximal neighborhood in the 781 claim. [00:08:19] Speaker 00: Then it applies a search query to the stack. [00:08:24] Speaker 00: And Van Vliet doesn't do that at all. [00:08:27] Speaker 00: Van Vliet is basically a system that stores and uses in various ways a user's browsing history. [00:08:35] Speaker 00: And there's a line in Van Vliet that talks about a functionality that uses the user history to basically filter out [00:08:45] Speaker 00: previously seen documents from the search results. [00:08:48] Speaker 00: So MED does not dispute, and I think this is clear from the appellate briefing, that unlike our patent claim, which excludes previously seen documents at this sort of initial stack building, neighborhood building step, Van Fleet [00:09:04] Speaker 00: simply hides results after the search query is done. [00:09:10] Speaker 00: And the board basically said, that's fine. [00:09:15] Speaker 00: We can take this very generalized teaching from Van Vliet that deals with a post query basically [00:09:23] Speaker 00: result filter and apply it way back here at the proximal neighborhood building step of the 781 zones. [00:09:34] Speaker 03: Is it disputed or undisputed whether the proximal neighborhood is what is initially searched through or what you ultimately [00:09:52] Speaker 03: decide is even post-filtering the result. [00:09:58] Speaker 03: I thought, is it? [00:09:59] Speaker 00: I believe it's undisputed in the 7.81 that the proximal neighborhood is what you search for. [00:10:05] Speaker 00: The system does cross a link of, cross links in a user's history to build the proximal neighborhood. [00:10:15] Speaker 00: And the proximal neighborhood is literally searched. [00:10:17] Speaker 03: I think I probably asked my question [00:10:20] Speaker 03: Badly so the last element is query, right? [00:10:24] Speaker 03: I was putting that aside you were distinguishing then fleet on the ground I think that it does some initial collection and then filters out the scene stuff [00:10:38] Speaker 03: Is it clear that, well, why doesn't, and then the result of the filtering before you then pose questions, the queries do it, is only unseen stuff. [00:10:50] Speaker 00: So I think Van Blee's a little simpler than that. [00:10:53] Speaker 00: It is only a post-query filter. [00:10:56] Speaker 00: It only takes seen stuff out after the query. [00:10:59] Speaker 00: There's a search query. [00:11:01] Speaker 00: some results come out and Van Vliet just hides the stuff that's been seen. [00:11:05] Speaker 00: And that part is undisputed. [00:11:07] Speaker 00: So our first point is it's too much of a radical sort of abstraction or generalization to take that very different kind of teaching and apply it to our proximal neighborhood step. [00:11:21] Speaker 00: We don't think a person of ordinary skill would do that. [00:11:24] Speaker 00: And we think that's reinforced by the structure of Koenig, which [00:11:29] Speaker 00: which doesn't just search the stack of documents that it builds. [00:11:35] Speaker 00: It basically does its crawling and creates a stack of documents it believes are going to be of high interest to the user, but when it runs its query, [00:11:45] Speaker 00: It runs the query in both the results of the personal crawling and what it calls all documents D, which is the set of all documents. [00:11:54] Speaker 00: This is all in column 27 of CONIG, which is the personal search function that Meta relies on. [00:12:01] Speaker 03: And you think that is it a fair description of the combination that the board relied on in the alternative that you don't take the full app? [00:12:15] Speaker 03: the entirety of Koenig or the entirety of Van Vliet. [00:12:21] Speaker 03: You take only the parts of the two that produce what's claimed here, which is the querying of a database that's already been restricted to Unseen. [00:12:33] Speaker 00: Well, I think all the board did was it took one very generalized teaching from Van Vliet's post-query filter and [00:12:43] Speaker 00: added it to VanVleet's stack building step. [00:12:47] Speaker 00: The board's using VanVleet's structure, which is build a stack, then add in all documents to run a query in the stack. [00:12:57] Speaker 00: That's Konig's structure, and the board simply plucked [00:13:03] Speaker 00: the general notion of hiding results that had been seen before and applied it to just the stack building step of coding. [00:13:13] Speaker 00: And that's what it would do. [00:13:15] Speaker 03: With motivation being that it's not figured out. [00:13:20] Speaker 00: We believe in motivation. [00:13:21] Speaker 00: Right. [00:13:22] Speaker 00: We believe the motivation's lacking. [00:13:23] Speaker 00: We think, as we say in our brief, we think code, if anything, teaches away from adding that kind of filter because code expressly teaches in column 27 the documents the user has seen before are of high interest. [00:13:37] Speaker 00: And again, it would be somewhat of ordinary skill would not, even if they were going to apply Van Vliet's filter, they wouldn't do it at the stack building step. [00:13:51] Speaker 00: because it would be pointless, because Konig searches in more than just its stack anyway. [00:13:57] Speaker 00: It searches in the stack and all documents D, so if you applied Van Vliet's filter to just the stack, the documents that are filtered out could [00:14:08] Speaker 00: be reincorporated into the search results, because more than just the stack is searched. [00:14:13] Speaker 00: If the person of ordinary scale was going to apply that teaching at all becoming, they would do it just the way Van Vliet does it, which is as a post-search filter. [00:14:23] Speaker 00: And if a person of ordinary scale did it that way, it wouldn't teach the claim. [00:14:27] Speaker 01: OK. [00:14:27] Speaker 01: You're into your rebuttal, so I'm here from the other side. [00:14:30] Speaker 01: And Mary Storson. [00:14:31] Speaker 01: Thank you. [00:14:31] Speaker 01: Thank you. [00:14:38] Speaker 02: Good morning. [00:14:40] Speaker 02: I'd like to start where we just left off, which is that's actually not the combination that was proposed. [00:14:47] Speaker 02: It was not a combination where all of a sudden you have to do all of the teachings of Van Vliet with all of the teachings of Konig. [00:14:53] Speaker 02: In fact, this is a substantial evidence case, right? [00:14:56] Speaker 02: So we look at what the board found, and what the board found, which was how we mapped it, is on Appendix 13, where the board specifically points out, and this is in the first full paragraph about halfway down, [00:15:09] Speaker 02: This characterization of van Vliet, however, is consistent with petitioner's contentions, which simply rely on van Vliet's teachings of restricting a search to items not previously viewed by a user, citing back to the petition. [00:15:25] Speaker 02: The paragraph goes on, and I think the most important is the last sentence, which says, [00:15:30] Speaker 02: Petitioner does not assert that Van Vliet alone teaches the proximal neighborhood subject matter of claim one. [00:15:38] Speaker 02: In fact, we strongly believe, as did the board below with substantial evidence, that Konig alone shows that only unseen documents come into the proximal neighborhood. [00:15:50] Speaker 02: The only reason that we added Van Vliet was if someone were to make the challenge, as they did, that the word only doesn't appear in Konig. [00:15:59] Speaker 02: And so one of ordinary skill in the art would look to see that there may in fact be reasons that you would want to restrict to only unseen documents, and that's the quote of Van Vliet that actually restricts the items to items not previously viewed. [00:16:15] Speaker 03: And just remind me, the reason why somebody might want to do that is [00:16:19] Speaker 02: For example, they're more interesting. [00:16:21] Speaker 02: You've already seen it. [00:16:22] Speaker 02: It spends less of your time. [00:16:24] Speaker 02: The documents that you haven't seen would be of more interest to you, because you've already seen other documents. [00:16:29] Speaker 02: And so you don't want to have to waste your time calling through those again. [00:16:32] Speaker 02: It's just like if you were to ask your clerks to show you cases that you haven't read in the last five days, because those are going to be of more interest to you, and you'd rather spend your time with those. [00:16:43] Speaker 02: Something as simple as that. [00:16:44] Speaker 01: And that was laid out in the petition. [00:16:46] Speaker 02: That was, in fact, Your Honor, yes, including the Fox Declaration supporting that, which talked about the fact that unseen documents may be of far more use. [00:16:55] Speaker 02: And I believe that's in paragraph 137 of Fox, which is on, let me get you an appendix site, 137 of Fox is at 715 of the appendix. [00:17:10] Speaker 02: But Your Honor, I think it's also critically important for us to go back to the substantial evidence that supports the board's decision that Koenig alone deals with only unseen documents. [00:17:23] Speaker 02: Unseen and seen are direct opposites to each other. [00:17:28] Speaker 02: If you tell someone to apply something to unseen documents, [00:17:32] Speaker 02: you are necessarily telling them don't do this for unseen documents. [00:17:38] Speaker 02: And that's exactly what we have in the PONIG reference, which the board specifically found. [00:17:44] Speaker 02: And this is at Appendix 12 of the board's decision. [00:17:47] Speaker 02: With the board acknowledging that you're trying to build a neighborhood of documents that are of interest to the user, the board first points to, so this is Appendix 12 in the first paragraph, [00:18:00] Speaker 02: They lay out what the patent owner's arguments are and then say they disagree. [00:18:04] Speaker 02: And then why they disagree. [00:18:06] Speaker 02: They don't just say they disagree. [00:18:07] Speaker 02: They explain why. [00:18:08] Speaker 02: And they point to the fact that CONIG starts with the fact that the term PUD represents the user interest in the document, regardless of the current information need, and is calculated using a user model. [00:18:23] Speaker 02: So then we look to see, OK, well, what is this user model? [00:18:27] Speaker 02: And the board goes on to say, [00:18:28] Speaker 02: that Konig states that each linked document is scored using the user model in step 176 of figure 18, and that that user model is applied to unseen documents to determine the probability that a document is of interest to the user. [00:18:45] Speaker 02: And that's in, as you've just noted, a section of the Konig reference at appendix 914 entitled Applying the User Model to Unseen Documents. [00:18:58] Speaker 02: The only analogy I could come up with that was as dramatically opposite would be if you were to say something along the lines of, give the food to the living people. [00:19:09] Speaker 02: That necessarily means you're not giving the food to the deceased people, something that dramatically different. [00:19:14] Speaker 02: So if you're applying it to unseen, you're not applying it to seen. [00:19:18] Speaker 02: And the board had some stamp letters. [00:19:19] Speaker 03: Can I just ask? [00:19:22] Speaker 03: If one had some doubts about the strength of the term necessary, does that change what we decide here? [00:19:33] Speaker 03: That is, if one thinks, well, in some context, that might be an implication, but not necessarily. [00:19:41] Speaker 02: So I think there's two ways to answer that, Your Honor. [00:19:43] Speaker 02: Actually, there's three ways. [00:19:44] Speaker 02: The first way is, no, I don't think it changes it, because the board only has to have substantial evidence to support its decision. [00:19:50] Speaker 02: And it does on page 12. [00:19:52] Speaker 02: And it lays it out very specifically. [00:19:55] Speaker 02: But if Your Honor is saying, [00:19:56] Speaker 02: Well, maybe somebody could argue that unseen doesn't always mean unseen. [00:20:01] Speaker 02: That's the reason to add in Van Vliet, which says we're going to restrict it to unseen. [00:20:07] Speaker 02: So there is no possibility of something else coming in. [00:20:10] Speaker 02: And that is also supported by the board's decision, and there's substantial evidence to support that fact as well. [00:20:17] Speaker 02: later down on Appendix Page 12, this time citing to Van Vliet at 954 and 955. [00:20:26] Speaker 02: This also, Your Honor, I think the third thing, I said there were three ways to answer your question. [00:20:31] Speaker 02: The third way to answer your question is to look to the HP versus Musetech and Parker Vision versus Qualcomm case, which [00:20:38] Speaker 02: points out that, as your honors have pointed out in numerous other cases, if there is one embodiment that does teach what you need to teach for obviousness, it does not matter that there may be other things that teach something else. [00:20:54] Speaker 02: As long as the teaching is there, that is enough. [00:20:57] Speaker 02: And the substantial evidence that the board relied on is set out very cleanly and very plainly on Appendix Page 12. [00:21:04] Speaker 02: And noting that none of the other arguments were raised today, unless your honors have questions, I would cede my time. [00:21:11] Speaker 02: Thank you. [00:21:12] Speaker 02: I appreciate your time and attention. [00:21:13] Speaker 02: Thank you. [00:21:22] Speaker 00: I'll respond to just a couple of points in the very limited time I have. [00:21:25] Speaker 00: First, the board, in our view, read only into the claim. [00:21:34] Speaker 00: The board also failed to consider contrary evidence, like the prosecution history I pointed to a moment ago, showing that Koenig is about ability, not excluding seen documents. [00:21:48] Speaker 03: Third, the prosecution history of the path of Koenig. [00:21:52] Speaker 00: Of Koenig. [00:21:54] Speaker 00: Of Koenig. [00:21:55] Speaker 00: the part where it says you don't distinguish between seen and unseen. [00:21:58] Speaker 00: And we believe that distinction is critical. [00:22:01] Speaker 00: Third, the case law that my opponent just referred to doesn't say that a reference can render something obvious just because it coincidentally may align with what the patent does. [00:22:17] Speaker 00: It would have to actually be configured to do that, at least in some environment which we believe is not met here. [00:22:26] Speaker 00: Thank you. [00:22:26] Speaker 00: We thank both sides for cases submitted.