`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`UNITED STATES DISTRICT COURT
`
`
`NORTHERN DISTRICT OF CALIFORNIA
`
`MASTEROBJECTS, INC.,
`
`
`
`Plaintiff,
`
`v.
`
`AMAZON.COM, INC,
`
`No. C 20-08103 WHA
`
`
`
`ORDER RE SUMMARY JUDGMENT
`
`Defendant.
`
`
`
`
`
`
`
`
`
`
`INTRODUCTION
`
`In this patent infringement action, defendant moves for summary judgment of
`
`noninfringement. To the following extent, the motion is GRANTED.
`
`STATEMENT
`
`This litigation concerns autocomplete technology for digital searches. Autocompletion
`
`suggests ways for the user to complete her search as she actively types it into a search bar.
`
`Patent owner MasterObjects, Inc. accuses alleged infringer Amazon.com, Inc. of infringing
`
`three of its patents in this area, U.S. Patent Nos. 9,760,628; 10,311,073; and 10,394,866.
`
`Our parties have some history. MasterObjects first filed a patent-infringement lawsuit
`
`against Amazon in 2011, but the parties stipulated to dismissal without prejudice less than six
`
`months later. MasterObjects, Inc. v. Amazon.com, Inc., No. C 11-01055 CRB (N.D. Cal. filed
`
`Mar. 7, 2011) (Judge Charles R. Breyer). MasterObjects targeted Amazon once again in May
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 2 of 17
`
`
`
`2020, when it initiated the instant lawsuit in the United States District Court for the Southern
`
`District of New York. In October 2020, Judge P. Kevin Castel transferred the suit to our
`
`district (Dkt. No. 82). MasterObjects originally alleged infringement of four patents, but one
`
`was later withdrawn.
`
`The three patents still in suit all descend from U.S. Patent No. 8,112,529 (filed in 2001).
`
`The following diagram lays out the patent genealogy. The asserted patents are highlighted in
`
`blue, full lines demarcate continuations of the earlier patent, and dashed lines represents
`
`continuations-in-part:
`
`The ’628 and ’866 patents are continuations of the original ’529 patent and share its
`
`specification. The ’073 patent is a continuation-in-part of the ’529 patent and incorporates its
`
`specification by reference.
`
`Of the fifteen claims asserted, there are three independent claims from which the rest of
`
`the claims-in-suit depend: claim 13 of the ’628 patent; claim 1 of the ’073 patent; and claim 1
`
`of the ’866 patent. To frame the analysis that follows, here is claim 13 of the ’628 patent in its
`
`entirety (emphasis added):
`
`13[a] A system comprising:
`
`13[b] a server system, including one or more computers, which is
`configured to receive query messages from a client object, the
`server system receiving and asynchronously responding to the
`query messages from the client object over a network;
`
`2
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 3 of 17
`
`
`
`13[c] the client object that, while a user is providing input
`comprising a lengthening string of characters, sends query
`messages to the server system;
`
`13[d] Whereby the query messages represent the lengthening
`string as additional characters are being input by the user; and
`
`13[e] wherein the server system, while receiving said query
`messages, uses the input to query data available to the server
`system and send return messages to the client object containing
`results in response to the input;
`
`13[f] wherein the server system caches query results and
`subsequently determines results by looking up the query in said
`cache so that it can avoid performing a query for the same input
`on a data source or looking up said query in a second cache.
`
`Per the common specification, “[a]s a user inputs data into a field on a form, the auto-complete
`
`function analyzes the developing character string and makes intelligent suggestions about the
`
`intended data being provided. These suggestions change dynamically as the user types
`
`additional characters in the string” (’628 patent 6:44–48). All agree that the claimed system
`
`generates autocomplete results from a specific server system as opposed to a broader search of
`
`the internet, generally.
`
`Of principal concern here, the key limitation in the claim recited above is directed to the
`
`system’s use of a “cache” to provide autocomplete results. “Cache” is used as both noun and
`
`verb. All the other claims-in-suit similarly reference a cache as well as a “data source” (or
`
`“content sources”). There is no dispute that “content sources,” like a “data source,” generally
`
`provide data to the claimed system. The lay reader will likely have a general awareness of a
`
`cache in the computing sense, such as a web browser (e.g., Mozilla Firefox) storing an iteration
`
`of a previously visited webpage (e.g., cand.uscourts.gov). This order addresses the use of the
`
`term “cache” by the patents-in-suit.
`
`Before going further into the claims, this order will get into the nuts and bolts of the
`
`accused system. Let’s start with a practical example of Amazon’s autocomplete as a user
`
`would find it on the Amazon.com homepage:
`
`3
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 4 of 17
`
`
`
`A user here has typed “gol” into Amazon’s search bar (an autocomplete query), and Amazon’s
`
`system has suggested several autocomplete query results, i.e., suggestions for a complete query
`
`for products such as “golf balls,” “gold standard whey protein,” etc. The autocomplete query
`
`results adjust as the user continues typing and revises her autocomplete query.
`
`Amazon’s autocomplete system uses a freestanding, constructed set of databases built by
`
`a process Amazon calls the “daily build.” The “daily build” process occurs every one or two
`
`days where it affirmatively builds two paired databases called read-only databases (RODBs).
`
`The RODB databases are the only part of Amazon’s system that can process an autocomplete
`
`query, an important point for noninfringement. These RODBs are, by definition, “read only”
`
`— data can neither be added nor removed from the databases during their one-to-two-day
`
`lifespan. The paired RODBs work together and store data in key-value pairs, as follows. The
`
`first RODB, the “Prefix RODB,” includes queries (i.e., prefixes) a user might type (e.g.,
`
`“gold,” “golde,” “golden”), paired with information reflecting rows in the second RODB. The
`
`second RODB, the “Keyword RODB,” stores the autocomplete suggestions (i.e., keywords)
`
`that correspond to the user’s query/prefix:
`
`While a user types, the autocomplete system finds a matching prefix in the Prefix RODB, if
`
`any, then follows the index numbers to the Keyword RODB, which the system then pushes to
`
`4
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 5 of 17
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`the user (Turnbull Noninfringement Rep. ¶¶ 103, 106–12, Dkt. Nos. 380-12, 384-5; Miller
`
`Decl. Exh. F, Dkt. No. 380-14).
`
`The key point, however, is how the columns are created in the first place. The RODBs
`
`are created from the “Tommy Query Groups.” Tommy is the name for the Amazon search
`
`dataset, which includes information on searches, clicks, purchases, etc. The Tommy Query
`
`Groups are aggregated Tommy search-analytics data related to completed product searches.
`
`Tommy Query Groups contain completed search strings from product search requests and
`
`metadata associated with each search request. This bears emphasizing because the Tommy
`
`dataset does not store autocomplete queries and at most stores autocomplete results (and
`
`arguably not even that). Rather, the Tommy dataset stores completed product search queries,
`
`i.e., not a query where the results are suggestions for a complete query (e.g., “Rol” = “Rolex”),
`
`but a query where the corresponding results are listings for actual products Amazon has for
`
`sale (e.g., the range of Rolex watches a customer could buy).
`
`The “daily build” process uses data from the Tommy Query Groups to populate the
`
`Keyword RODB (the results column). During this process, among other things, spelling-
`
`corrected suggestions are generated and a blocklist is applied to remove certain entries. Once
`
`the “daily build” collects all the keywords in the Keyword RODB, it generates prefixes for
`
`each using a rote, progressive letter construction: for a keyword such as, e.g., “Rolex,” it
`
`would generate the prefixes of “R,” “Ro,” “Rol,” “Role,” and “Rolex.” Amazon also uses a
`
`“Personalization Platform Data Service” (PPDS), which includes the individual user’s recent
`
`completed product search queries, which are used to re-rank suggestions to more closely
`
`match those prior queries (see Turnbull Noninfringement Rep. ¶¶ 109–12, 171, 181 n.21, 244;
`
`Peck Rep. ¶¶ 140–41, Dkt. Nos. 380-6, 384-7, 389-2).
`
`Having completed this high-level overview of the patents-in-suit and the accused system,
`
`this order proceeds to analyze the specifics. This order follows full briefing and oral argument.
`
`ANALYSIS
`
`A patent infringement analysis involves two steps. The claim must be properly construed
`
`to determine its scope and meaning. Claim terms generally take “their ordinary and customary
`
`5
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 6 of 17
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`meaning,” that is “the meaning that the term would have to a person of ordinary skill in the art
`
`in question at the time of the invention.” Phillips v. AWH Corp., 415 F.3d 1303, 1312–13
`
`(Fed. Cir. 2005) (en banc). Although construction begins with the claim language itself, “the
`
`specification is the single best” — and usually dispositive — “guide to the meaning of a
`
`disputed term.” Network-1 Techs., Inc. v. Hewlett-Packard Co., 981 F.3d 1015, 1022 (Fed.
`
`Cir. 2020) (quoting Phillips, 415 F.3d at 1314–15).
`
`The properly construed claim must then be compared to the accused device or process.
`
`To constitute infringement, an accused product must practice every limitation of a properly
`
`construed claim. Tessera, Inc. v. Int’l Trade Comm’n, 646 F.3d 1357, 1364 (Fed. Cir. 2011);
`
`Carroll Touch, Inc. v. Electro Mech. Sys., Inc., 15 F.3d 1573, 1576 (Fed. Cir. 1993).
`
`1.
`
`AMAZON’S SYSTEM DOES NOT HAVE A CACHE OF PRIOR
`AUTOCOMPLETE QUERIES, AS REQUIRED BY THE CLAIMED
`SYSTEM.
`
`Amazon asserts the claims-in-suit all require a cache that stores prior autocomplete
`
`queries. MasterObjects opposes that the claimed system only requires prior autocomplete
`
`results and that Amazon’s construction reads limitations from the embodiment into the claims.
`
`The crux of the dispute is the construction of “cache.” Because this order construes the
`
`claimed system’s “cache” to store autocomplete queries (as well as results), Amazon’s system
`
`does not infringe.
`
`All of the claims-in-suit include “cache and content source” terms. The three
`
`independent claims-in-suit recite:
`
`• “[T]he server system caches query results and subsequently determines results by
`
`looking up the query in said cache so that it can avoid performing a query for the
`
`same input on a data source or looking up said query in a second cache” (’628
`
`patent claim 13[f]).
`
`• “[M]atching, by the server system, the string [representing an incomplete search
`
`query] to entries in a cache of query strings and search results based on content
`
`queries received from multiple users, whereby cached search results contain a
`
`subset of data from one or more content sources” (’073 patent claim 1[e]).
`
`6
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 7 of 17
`
`
`
`• “[M]atching, by the server system, the string [representing an incomplete version of
`
`the search query] to entries in a cache of queries and search results previously
`
`retrieved from one or more content sources” (’866 patent claim 1[e]).1
`
`In their claim construction statements, Amazon construes each of these claim elements as
`
`a whole, while MasterObjects construes various component terms. For example, here are the
`
`proposed constructions for claim 13[f] of the ’628 patent (the proposed constructions for the
`
`other two cache limitations are similar):
`
`MasterObjects
`
`Amazon
`
`“the server system saves a copy of the
`query result set retrieved in response to
`the query message and determines
`responses to later query messages by
`checking whether result sets responsive
`to those later query messages are
`present in its store of previously saved
`query result sets before querying a
`separate data source or a second cache”
`
`This claim requires that the claimed
`system be able to obtain query result sets
`from sources other than the store of
`query result sets previously saved by the
`server system.
`
`• “Caches query results:” saves query
`results through a cache;
`
` •
`
` “Cache:” a memory store;
`
`• “Data Source:” an underlying data
`source;
`
` •
`
` •
`
` The remainder: plain and ordinary
`meaning; and
`
` This term is not limited to a system
`that can, if no matching cache entry
`is found, query another content
`source
`
`The use of the word “cache” by itself begs the question of what must be stored in the
`
`cache. A cache, in common parlance, is simply a fast memory buffer for holding frequently
`
`called data (see Webster’s Pocket Computer Dictionary, Turnbull Claim Const. Resp. Exh. 29;
`
`see also id. Exh. 28). What must be stored (autocomplete queries and results or just results?) is
`
`the question.
`
`All three independent claims indicate that, a “cache,” in context, is not simply “a memory
`
`store.” First, claim 13 specifies that the system “caches query results,” but goes on to explain
`
`that the system “look[s] up the query in said cache.” The system’s ability to look up a query in
`
`
`1 This order has labelled each individual claim limitation, e.g., claim 13[f], for ease of reference.
`7
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 8 of 17
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`the cache inherently indicates the system has also stored the prior query as well. Furthermore,
`
`as indicated by “subsequently” and “avoid,” the use of the verb form of cache expresses that
`
`data stored in the cache had previously been processed by the data source. Second, claim 1[e]
`
`of the ’073 patent similarly recites that the claimed system has a “cache of query strings,” i.e.,
`
`store of autocomplete queries. The limitation describes the claimed system’s cache as being “a
`
`subset of data from one or more content sources,” where that data is “based on content queries
`
`received from multiple users.” This description of the cache also evinces that it stores
`
`autocomplete queries already processed by a data source. Third, claim 1[e] of the ’083 patent
`
`explicitly recites that the system has a “cache of queries and search results” that have been
`
`“previously retrieved from one or more content sources.” This limitation makes clear the
`
`cache stores autocomplete queries previously processed by a content/data source.
`
`Upon review, this order construes a “cache” as a particular type of memory store that
`
`saves a copy of both the autocomplete query and results from the data source. This order
`
`likewise construes the verb form of “cache” to mean to store a copy of both the autocomplete
`
`query and results from the data source.
`
`Shifting sands have blown hard in this case. MasterObjects began this litigation
`
`expressly in agreement that the claims required a cache of both autocomplete queries and
`
`results. As this litigation progressed, however, it has become clear that the accused system
`
`does not store queries, so MasterObjects flip-flopped in its opposition to summary judgment,
`
`reneging on its prior assertions that the cache terms did indeed require storage of autocomplete
`
`queries. For example, in MasterObjects’ claim construction briefing:
`
`Amazon begins by misdescribing MasterObjects’ proposed
`definition of the caching terms . . . . That is, MasterObjects [sic]
`claimed cache is based on prior search queries (e.g., “ROL”)
`with prior query results (e.g., “Rolex watches”). This is not a
`subtle point in MasterObjects’ briefing
`
`(MO Claim Const. Reply Br. 5, Dkt No. 262 (emphasis added); see also MO Claim Const. Br.
`
`1–2, Dkt. No. 236 (same); id. at 22 (same); Fourth Supp. Infringement Contentions (FSIC) at
`
`78, Dkt. No. 388-8 (same)). MasterObjects went on, discussing its cache construction in other
`
`cases:
`
`8
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 9 of 17
`
`
`
`But the real MasterObjects construction here accords with its prior
`construction: (1) Meta: “cache” is “a store that includes previous
`queries and search results retrieved in response to previous
`queries;” (2) eBay: “query and result cache” is “[a] cache which
`stores previous queries and content or other information returned
`in response to the previous queries;” and (3) Yahoo!: “query and
`result cache” is “[a] cache which stores previous queries and
`content or other information returned in response to the previous
`queries.” These are vastly contradictory constructions?
`
`(MO Claim Const. Reply Br. 5, emphasis in original).2
`
`MasterObjects’ Expert Peck stated the same conclusion multiple times in his
`
`infringement report:
`
`• “AC [Amazon’s autocomplete system] does so by using multiple pre-computed
`
`query results in “Query and Results Caches” in its server-side backend. Generally,
`
`a Query and Results Cache is a memory store of prior queries and search results”
`
`(Peck Rep. ¶ 120).
`
`• “The 13[F] claim language describes a process where: (1) a server system caches
`
`query results, which inherently includes the prior queries themselves (as one would
`
`not know what a query result was absent having the query in question) . . . .” (id. at
`
`¶ 198).
`
`• “In the language of the specification and claims, [claim 13] requires that Amazon
`
`build a cache of prior queries and results, e.g., ‘ROL’ likely means Rolex. . . .
`
`[T]he cache is built and designed to correlate prior queries (which can be query
`
`fragments, e.g., ‘ROL’) with the list of most likely relevant results” (id. at ¶ 199).
`
`The reader should please note this is MasterObjects’ current expert infringement report
`
`disclosed in advance of trial. This order agrees with the original and persistent claim
`
`construction proffered by MasterObjects itself.
`
`
`2 Citing MasterObjects, Inc. v. Meta Platforms, Inc., No. C 21-05428 WHA (N.D. Cal.);
`MasterObjects, Inc. v. Yahoo!, Inc., 2013 WL 6185475 (N.D. Cal. Nov. 26, 2013) (Judge Jeffrey
`S. White); MasterObjects, Inc. v. eBay, Inc., 2013 WL 1287428 (N.D. Cal. Mar. 28, 2013) (Judge
`Jacqueline Scott Corley). See also MasterObjects, Inc. v. Google, Inc., 2013 WL 2319087 (N.D.
`Cal. May 28, 2013) (Judge Phyllis J. Hamilton).
`
`9
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 10 of 17
`
`
`
`Other intrinsic evidence supports this construction, such as the inter-partes review
`
`proceeding for the ’073 patent. In that proceeding, MasterObjects’ Expert Dr. Michael J.
`
`Pazzani opined regarding claim 1[e]:
`
`Even if the predictive search server [of the prior art] stores that
`information that does not mean it is necessarily stored in a cache.
`A POSITA would understand that all data is stored in memory to
`make it usable by system [sic]. Merely storing data in memory
`(without more) does not require or even suggest to a POSITA that
`the data is stored in a cache, at least because the data stored must
`include “previous queries and search results,” and those results
`must have been “retrieved in response to previous queries”
`
`(Amazon Claim Const. Resp. Exh. 35, Dkt. No. 252-57). Dr. Pazzani expressly states the
`
`claim element requires storage of previous queries.
`
`The common specification also accords with the construction this order has adopted.
`
`MasterObjects now decries analysis of the specification and asserts that Amazon “commits
`
`perhaps the cardinal sin of construction errors — importing embodiment detail into the claim
`
`language.” The common specification describes one preferred embodiment of the invention,
`
`“QuestObjects.” MasterObjects says the embodiment is “replete with odd neologisms” that do
`
`not appear in the claims (MO Claim Const. Reply Br. 1, emphasis in original).
`
`This order cannot, however, simply ignore the specification — the specification “is
`
`always highly relevant to the claim construction analysis.” Phillips, 415 F.3d at 1315 (citation
`
`omitted). The common specification states: “to ensure scalability, systems built around the
`
`present invention can be divided into multiple tiers, each tier being capable of caching data
`
`input and output” (’628 patent 6:16–18, emphasis added). This description is discretionary
`
`(“can”), but it provides insight into how to read the cache limitations. Furthermore, this
`
`statement in the specification is made before the introduction of the QuestObjects embodiment
`
`(id. at 7:24–32). Accordingly, the reference to “the present invention” indicates the description
`
`refers to the invention as a whole rather than only the preferred embodiment. See Verizon
`
`Servs. Corp. v. Vonage Holdings Corp., 503 F.3d 1295, 1308 (Fed. Cir. 2007).
`
`The QuestObjects “Glossary” also defines a “Content-based Cache” as a “persistent store
`
`of Queries and corresponding Result Sets executed by a Content Engine for a specific Content
`
`10
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 11 of 17
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Channel” (id. at 10:36–38). (A “Content Engine,” per QuestObjects, is a data source, while a
`
`“Content Channel” provides one type of information from one Content Engine (id. at 10:39–
`
`45).)
`
`Yes, QuestObjects is an embodiment. But this order merely takes note of the fact that the
`
`specification accords with the other intrinsic evidence regarding the cache limitations. Take,
`
`for example, Judge Phyllis Hamilton’s claim construction order in MasterObjects’ prior
`
`litigation against Google. There, Google pointed to the same “Glossary” definition cited above
`
`and tried to incorporate the word “persistent” into the construction of “content-based cache”
`
`and “query and result cache.” Judge Hamilton noted MasterObjects’ inconsistency by
`
`distancing itself from the “Glossary” definition, but found including “persistent” would create
`
`more ambiguity, not less. See Google, 2013 WL 2319087, at *4. This order cites the
`
`“Glossary” definition to highlight the uniform description of the cache limitations in the
`
`intrinsic evidence. “Where, as here, a patent repeatedly and consistently characterizes a claim
`
`term in a particular way, it is proper to construe the claim term in accordance with that
`
`characterization.” Wis. Alumni Research Found. v. Apple Inc., 905 F.3d 1341, 1351 (Fed. Cir.
`
`2018) (citations and quotations omitted); see also Decisioning.com, Inc. v. Federated Dep’t
`
`Stores, Inc., 527 F.3d 1300, 1307–08, 1311 (Fed. Cir. 2008); Phillips, 415 F.3d at 1323.
`
`MasterObjects now tries to walk back its prior statements with the excuse that its “prior
`
`vernacular has not always been perfect” (Opp. 16). But its citations to other passages in Expert
`
`Peck’s reports to support its current position are either inapposite descriptions of Amazon’s
`
`system or general statements that are consistent with Expert Peck’s specific statements
`
`highlighted above (ibid., citing Peck Rep. ¶¶ 153–56; Peck Invalidity Rebuttal Rep. ¶¶ 44, 46,
`
`50, Dkt. No. 383-6).
`
`MasterObjects now primarily relies on its proposed plain-meaning constructions for
`
`various parts of the cache limitations to argue that it has consistently asserted that only prior
`
`autocomplete results are required to practice the claims-in-suit. Claim 1[e] of the ’866 patent
`
`for example, describes in relevant part “a cache of queries and search results previously
`
`retrieved from one or more content sources.” MasterObjects argues the phrase “previously
`
`11
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 12 of 17
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`retrieved from one or more content sources” only modifies “search results,” not “queries”
`
`(Opp. 14–15). So, according to MasterObjects, the plain language does not require the cache
`
`to store prior autocomplete queries. But reaching this conclusion reads “cache” out of the
`
`claims and untethers the query results from the query itself. Construed in the proper context, a
`
`“cache of queries and search results” indicates that the cache stores an autocomplete query
`
`along with the corresponding results generated from processing of that query. MasterObjects’
`
`argument that autocomplete queries come from the user rather than the backend server is a
`
`strawman. The point is that autocomplete queries that the system processes to generate results
`
`are stored in the cache because (in the words of MasterObjects’ expert), “one would not know
`
`what a query result was absent having the query in question” (Peck Rep. ¶ 198). See Irdeto
`
`Access, Inc. v. Echostar Satellite Corp., 383 F.3d 1295, 1300 (Fed. Cir. 2004).
`
`MasterObjects’ “best” argument is as follows: even though the autocomplete query is
`
`not stored in a “cache,” the accused system generates a sequence of prefixes — one of which
`
`by definition must correspond exactly to the query — and stores them in a “cache.” More
`
`specifically, for example, compare these: under the patented system, a user types the
`
`autocomplete query “Rol” and one of the query results will be “Rolex.” It then stores both
`
`“Rol” and “Rolex” in respective columns in its cache. Pursuant to this argument, under
`
`Amazon’s system a user types the autocomplete query “Rol” and one result is “Rolex.”
`
`“Rolex” is then stored in a “cache.” (More on what is actually stored in a moment.) The
`
`accused system then uses a progressive letter construction to store the following: “R,” “Ro,”
`
`“Rol,” “Role,” and “Rolex.” All of these are stored in the “cache” as well. The main problem
`
`with this argument is the autocomplete query itself is not stored and the prefix column in the
`
`“cache” is not a history of prior queries. Rather, the Prefix RODB is an artificially generated
`
`set of all possible prefixes that would lead to “Rolex.” “Rol” in the accused system is not a
`
`copy of the autocomplete query but an independently and artificially generated construct. This
`
`is sufficient for a finding of noninfringement.3
`
`
`3 Amazon’s system could also hypothetically have the prefix “Role” for the autocomplete result
`“Rolex” even though no prior user had ever actually typed in the autocomplete search “Role,”
`12
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 13 of 17
`
`
`
`Furthermore, setting aside that the claims in-suit require caching autocomplete queries,
`
`Amazon’s system also does not infringe because the Keyword RODB is built from prior
`
`product search queries, not prior autocomplete results like the claimed system. MasterObjects
`
`argues that a product search query selected by a user from an autocomplete result is an
`
`autocomplete result because the query run comes verbatim from the autocomplete result, e.g.,
`
`both would be “Rolex” (Opp. 5). The problem with this theory is that Amazon processes
`
`product search queries with an entirely different search system located on its own separate
`
`servers (Turnbull Noninfringement Rep. ¶¶ 125–26). Amazon’s system may use information
`
`gleaned from a user’s prior actions to help generate autocomplete results, but it gathers that
`
`data from an entirely different search system (for product searches). And MasterObjects’
`
`claimed system specifically caches autocomplete data. That is not the same. The same
`
`reasoning applies for why the PPDS does not qualify as a cache. (Regardless, the PPDS also
`
`cannot be queried and does not otherwise push autocomplete results to the user; it is only used
`
`to re-rank an individual user’s autocomplete results.) Consequently, there is no infringement
`
`on this basis as well.
`
`2.
`
`AMAZON’S SYSTEM DOES NOT HAVE AN UNDERLYING DATA
`SOURCE THAT CAN ITSELF BE QUERIED, AS REQUIRED BY THE
`CLAIMED SYSTEM.
`
`The reason that MasterObjects painted itself into a corner regarding the claimed system’s
`
`caching of prior autocomplete queries has to do with the parties’ dispute over the overarching
`
`structure of the claimed system. As explained, a “cache” is a particular type of memory store
`
`that saves a copy of both the autocomplete query and results from the data source. This
`
`inherently requires the claimed system to be able to query the data source (or “content
`
`sources”) for autocomplete query results. Because the “Tommy Query Groups” dataset that
`
`the accused system uses to build the RODBs cannot itself be queried for autocomplete results,
`
`Amazon does not infringe.
`
`
`which could not happen for MasterObjects’ claimed system.
`13
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`Northern District of California
`
`United States District Court
`
`
`
`Case 3:20-cv-08103-WHA Document 415 Filed 09/15/22 Page 14 of 17
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`26
`
`27
`
`28
`
`The analysis of the claim language above indicated not only that all three cache
`
`limitations required caching of prior autocomplete queries, but also that the claimed system has
`
`an underlying data source that itself can be queried. As just one illustration to reorient the
`
`reader, claim 13[f] of the ’628 patent recites a system that “can avoid performing a query for
`
`the same input on a data source.” This permissive language indicates the data source itself can
`
`be queried if need be.
`
`The common specification accords with this construction. It says, prior to the
`
`introduction of the QuestObjects embodiment: “Using the present invention, the suggestions
`
`generated by the server may, at the option of the application developer, be cached on the
`
`middle tier or on the client itself to maximize performance” (’628 patent 7:3–6). Since caching
`
`is a discretionary optimization, the claimed system must be able to query the underlying data
`
`source to generate autocomplete query results. See Verizon, 503 F.3d at 1308.
`
`This order also notes that the rest of the specification uniformly and repeatedly describes
`
`the data source as a part of the sy