`Patent No. 5,832,494
`
`Paper No. ___________
`
`
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`_________________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`_________________
`
`FACEBOOK, INC., LINKEDIN CORP., and TWITTER, INC.,
`Petitioners
`v.
`
`SOFTWARE RIGHTS ARCHIVES, LLC
`Patent Owner
`
`
`
`Case IPR2013-00480
`Patent No. 5,832,494
`
`
`
`PETITIONERS’ REPLY TO PATENT OWNER’S RESPONSE
`
`
`
`
`
`
`
`859473.01
`
`
`
`Table of Contents
`
`
`Page
`
`
`
`2.
`3.
`
`4.
`
`5.
`
`INTRODUCTION .......................................................................................... 1
`I.
`CLAIM CONSTRUCTION ........................................................................... 2
`II.
`III. CLAIMS 1 AND 5 ARE ANTICIPATED BY FOX SMART ...................... 3
`A.
`Fox SMART explicitly discloses each element of Claim 1 ................. 3
`1.
`Fox SMART discloses “analysis of one or more indirect
`relationships in [a] database” ..................................................... 3
`Fox SMART discloses “selecting a node for analysis” ............. 5
`Fox SMART discloses “generating candidate cluster
`links for the selected node” by analyzing “indirect
`relationships in the database” ..................................................... 5
`Fox SMART discloses “deriving actual cluster links from
`the candidate cluster links” ........................................................ 7
`Fox SMART discloses “identifying one or more nodes
`for display” and “displaying the identity of one or more
`nodes using the actual cluster links” .......................................... 8
`Fox SMART discloses the steps of claim 1 as arranged in the
`claim ..................................................................................................... 9
`Fox SMART discloses “eliminating candidate cluster links” by
`choosing the closest links, as required by Claim 5 .............................. 9
`IV. CLAIMS 14 – 16 ARE ANTICIPATED BY FOX THESIS ....................... 10
`A.
`Fox Thesis explicitly discloses each element of Claim 14 ................ 10
`1.
`Fox Thesis discloses “initializing a set of candidate
`cluster links” ............................................................................ 10
`Fox Thesis discloses “selecting the destination node of a
`path as the selected node to analyze” ....................................... 11
`Fox Thesis discloses “retrieving the set of direct links
`from the selected node to any other node in the database”...... 11
`Fox Thesis discloses “determining the weight of the path
`using the retrieved direct links” ............................................... 12
`
`B.
`
`C.
`
`2.
`
`3.
`
`4.
`
`
`
`-i-
`
`
`
`
`
`
`
`
`V.
`
`Fox Thesis discloses “repeating” steps 3 through 5 above
`“for each path” ......................................................................... 13
`Fox Thesis discloses “storing the determined weights as
`candidate cluster links” ............................................................ 13
`Fox Thesis discloses “deriving the actual cluster links” as a
`subset of the candidate links, as required by Claim 15 ...................... 13
`Fox Thesis discloses further “choosing the top rated candidate
`cluster links,” as required by Claim 16 .............................................. 14
`CONCLUSION ............................................................................................. 14
`
`B.
`
`C.
`
`5.
`
`6.
`
`
`
`-ii-
`
`
`
`
`
`
`
`
`I.
`
`INTRODUCTION
`
`Fox SMART (Exhibit 1005) and Fox Thesis (Exhibit 1008) anticipate claims
`
`1 and 5 and 14-16 of U.S. Patent No. 5,832,494 (“the ’494 patent”), the only
`
`claims that remain at issue in this proceeding.1 A decade before the applicants
`
`filed the parent application for the ’494 patent, those references disclosed
`
`clustering and clustered searching based on co-citation (cc) and bibliographic
`
`coupling (bc) relationships between documents in a digital database, exactly as
`
`claimed.
`
`Patent Owner argues that Fox SMART and Fox Thesis discussed
`
`relationships between “paper” documents only, not electronic ones, because the
`
`test collection that Dr. Fox used for his research (the CACM collection) contained
`
`only the abstract and other information for each article, not the full text. See Resp.
`
`1, 26. But this argument is irrelevant; nothing in the claim limitations requires
`
`textual objects in the database that cite to each other.
`
`Patent Owner raises other objections that argue irrelevant points, such as
`
`whether the claimed step is “required” or only disclosed in the prior art. See, e.g.,
`
`Resp. 41-42 (admitting that Fox SMART discloses that documents are displayed in
`
`
`1 After the Board instituted this inter partes review, Patent Owner cancelled
`
`challenged claims 8, 10, 11, 35, and 40. See Resp. 12. Petitioners do not oppose.
`
`
`
`-1-
`
`
`
`
`
`
`
`
`response to a user query, but suggesting that the system also had a “batch” mode).
`
`As explained in more detail below, none of Patent Owner’s objections refute the
`
`art’s clear teachings.
`
`II. CLAIM CONSTRUCTION
`For the purpose of this proceeding, Petitioners use the constructions adopted
`
`by the Board in its Institution Decision. See Paper 17 (“Inst. Dec.”) 9-13.
`
`Petitioners disagree with Patent Owner’s claim that the Board made a “technical
`
`error” when it construed “indirect relationships in the database.” See Resp. 21-23.
`
`The Board construed that term consistently with its construction of “indirect
`
`relationships” in the co-pending IPRs, because the ’494 patent contains
`
`continuation material that substitutes “links and nodes” language for the “citation
`
`relationship” language of the ’352 patent.2 But this issue is academic, because the
`
`2 The ’494 patent application was a continuation-in-part of the ’352 application,
`
`and its new matter included disclosure of using “links and nodes to index and
`
`search a database.” Compare ’494 Patent, Abstract, with ’352 Patent, Abstract.
`
`Thus, where the ’352 patent refers to citation relationships, the ’494 patent uses the
`
`more general language of links and nodes. See, e.g., ’494 Patent 51:38-39 (claim 1
`
`preamble: “indirect relationships, using links and nodes”) (emphasis added). It is
`
`therefore entirely appropriate that, in the ’494 patent, the Board construed “indirect
`
`relationships in the database” as relationships “characterized by at least one
`
`
`
`-2-
`
`
`
`
`
`
`
`
`challenged claims are anticipated under both the Board’s construction and Patent
`
`Owner’s alternative construction, which requires database objects to be connected
`
`by “a chain of citations.” Exhibit 1028 (“Fox Reply Decl.”) ¶¶ 269-70.
`
`III. CLAIMS 1 AND 5 ARE ANTICIPATED BY FOX SMART
`A.
`Fox SMART explicitly discloses each element of Claim 1
`
`
`1.
`
`Fox SMART discloses “analysis of one or more indirect
`relationships in [a] database”
`
`Fox SMART describes an algorithm for clustering documents for given
`
`concept types—including bc and cc, which represent indirect relationships. See
`
`Exhibit 1005 (“Fox SMART”) 44-46; Fox Reply Decl. ¶ 272. Fox SMART
`
`teaches analyzing the “raw data” CITED tuples to generate bc and cc subvectors
`
`and corresponding submatrices containing the indirect relationships in the
`
`database, including bibliographic coupling and co-citation. Fox SMART 30-32.
`
`Fox SMART explains that, since the 1960s, the SMART system was used to
`
`test new ideas “aimed at fully automatic document retrieval,” including, e.g., “full-
`
`text retrieval applications.” Fox SMART 1, 80. Published in 1983, Fox SMART
`
`teaches the use of extended vectors—including bibliographic relationships—with
`
`the SMART system, and the “manipulation of these extended vectors . . . [for]
`
`clustering and clustered searching[.]” Id. 2.
`
`
`intermediate node between two nodes.” Inst. Dec. at 11.
`
`
`
`-3-
`
`
`
`
`
`
`
`
`Thus, for example, Fox SMART teaches how to programmatically generate
`
`and update matrices of bc and cc relationships within a document collection where
`
`“only basic raw data is supplied.” Id. 29-35; Pet. 18; Fox Decl. ¶¶ 157-158. That
`
`raw data “will be a set of tuples describing which documents are cited by others[.]”
`
`Fox SMART 29. Fox SMART explains that, using the disclosed techniques,
`
`“given raw data such as that document 5734 refers to 3256 and 4681,” the system
`
`can derive a co-citation relationship between documents 3256 and 4681 and store
`
`that relationship in a vector. Id. 38. The subvectors bc and cc were stored in the
`
`database. Fox SMART 2, 5, 9, 13-21, 27-40, 80-81.
`
`Patent Owner argues that Fox SMART does not teach this limitation because
`
`“Fox SMART does not disclose or suggest that the bc and cc subvectors . . .
`
`represent citation relationships between objects in a database.” Resp. 26; see
`
`generally id. 23-29. Patent Owner further faults Fox SMART for not teaching that
`
`the contents of the documents’ bibliographies are stored in the database. Id. 26-27.
`
`Patent Owner’s argument is a red herring: nothing in claim 1 requires that the
`
`indirect relationships be between objects in a database. All that is required is the
`
`“analysis of one or more indirect relationships in [a] database,” and Fox SMART
`
`discloses that limitation. See Fox SMART 30-31, 44-46; Fox Reply Decl. ¶¶ 273-
`
`74. Thus, Fox SMART discloses “analysis of indirect relationships in [a]
`
`database” under both the Board’s claim construction and the Patent Owner’s.
`
`
`
`-4-
`
`
`
`
`
`
`
`
`
`2.
`
`Fox SMART discloses “selecting a node for analysis”
`
`Like virtually all clustering operations, Fox SMART discloses selecting a
`
`node for analysis. For example, Fox SMART describes a clustering process where
`
`documents are added one by one as “leaves” of a multilevel tree. Fox SMART 44;
`
`see Fox Reply Decl. ¶¶ 275-276.
`
`
`3.
`
`Fox SMART discloses “generating candidate cluster links
`for the selected node” by analyzing “indirect relationships
`in the database”
`
`As the Board found, Fox SMART teaches clustering based on analysis of
`
`indirect relationships. See Inst. Dec. 17; Fox SMART 30-32, 41-44, 46. The
`
`clustering algorithm disclosed in Fox SMART was generalizable and could be
`
`performed using a combination of direct and indirect relationships, or using
`
`indirect relationships alone. Fox Reply Decl. ¶ 278. The clustering algorithm of
`
`Fox SMART produces a hierarchy or classification in which all of the N
`
`documents in the collection end up as leaves of a multilevel tree. Fox SMART at
`
`44. As already explained above, Fox SMART discloses the construction of bc and
`
`cc submatrices which contain all of the bibliographic coupling and co-citation
`
`relationships in the database. Fox SMART at 30-32. Fox SMART further teaches
`
`clustering based on “given concept types” and “available subvectors.” Fox
`
`SMART 44-46 & Table 3. Bc and cc—which represent indirect relationships—are
`
`both concept types and available subvectors. Fox Reply Decl. ¶ 278.
`
`
`
`-5-
`
`
`
`
`
`
`
`
`Patent Owner appears to concede that Fox SMART teaches clustering based
`
`on analysis of indirect relationships. See Resp. 17 (stating that an artisan would
`
`understand that the “internal form,” which includes cc and bc, “is what is used for
`
`clustering, searching, etc.”). It argues only that Fox SMART did not require users
`
`to cluster based on indirect relationships, which is irrelevant. See Resp. 31-32.
`
`Fox SMART also discloses generating “candidate cluster links”—i.e., a set
`
`of possible cluster links between a search node and a target node—for the selected
`
`node based on an analysis of indirect relationships. Candidate cluster links are
`
`generated during the clustering operation before final clusters are determined. See
`
`Fox SMART 44, 46, 49-50. For example, when an oversize cluster is split into
`
`smaller clusters (an operation that is performed repeatedly as documents are added
`
`to the cluster tree), Fox SMART teaches that a pairwise similarity value is
`
`calculated for every node to every other node in the original cluster. See Fox
`
`SMART at 49-50; Fox Reply Decl. ¶ 279. Each pairwise similarity value is a
`
`candidate cluster link. Fox SMART teaches that these similarity values may be
`
`based on indirect relationships. See Fox SMART at 46 (“the overall similarity
`
`between documents can be determined based on available subvectors,” which
`
`include cc and bc); Fox Reply Decl. ¶ 279.
`
`After the pairwise similarity values are calculated, “candidate clusters” are
`
`formed from “highly correlated pairs,” and subjected to the “concentration,”
`
`
`
`-6-
`
`
`
`
`
`
`
`
`“definition,” and “uncour” (overlap) tests. Fox SMART at 48-51. These
`
`“candidate clusters” also comprise candidate cluster links. They, too, are based on
`
`analysis of indirect relationships because the “highly correlated pairs” are derived
`
`from the pairwise similarity values. Id.; Fox Reply Decl. ¶ 279.
`
`
`4.
`
`Fox SMART discloses “deriving actual cluster links from
`the candidate cluster links”
`
`This intuitive step requires only that the actual cluster links be a subset of
`
`the candidates that meet certain criteria. See Inst. Dec. 12-13. And, as the Board
`
`found, Fox SMART teaches this. See Inst. Dec. 18. For example, when an
`
`oversize cluster is split into smaller clusters, a pairwise similarity value is
`
`calculated for every node to every other node in the original cluster, each of which
`
`is a candidate cluster link. See Fox SMART at 49-51; Fox Reply Decl. ¶ 280. The
`
`clusters that are eventually formed comprise a subset of these candidate links that
`
`meet certain criteria—e.g., “having enough highly correlated pairs,” and otherwise
`
`satisfying the concentration, definition, and uncour tests. Id.; see also Fox
`
`SMART 51 (“Eventually, only clusters that pass all appropriate tests are
`
`accepted.”). Furthermore, the clustering algorithm described in Fox SMART
`
`produces a final clustering where there are multiple clusters where the remaining
`
`nodes have previously computed pairwise similarity values—in other words, the
`
`similarity values are candidate cluster links that have become actual cluster links.
`
`Fox Reply Decl. ¶ 280.
`
`
`
`-7-
`
`
`
`
`
`
`
`
`Patent Owner argues that the concentration, definition, and uncour tests do
`
`not “delete” any candidate clusters, see Resp. at 34, 37-38, but this is both wrong
`
`and irrelevant. Fox SMART teaches that potential (i.e., candidate) clusters are
`
`rejected (“deleted”) if they fail any one of these tests. See Fox SMART 51
`
`(“Eventually, only clusters that pass all appropriate tests are accepted.”); Fox
`
`Reply Decl. ¶ 281. In any event, this claim limitation does not require that
`
`candidate clusters be “deleted,” only that the actual cluster links be a subset of the
`
`candidates that meet certain criteria.
`
`
`5.
`
`Fox SMART discloses “identifying one or more nodes for
`display” and “displaying the identity of one or more nodes
`using the actual cluster links”
`
`As the Board found, Fox SMART discloses searching a clustered tree of
`
`documents so that the documents in a “retrieved cluster are presented to the user.”
`
`See Inst. Dec. 18; Fox SMART 53-54. Patent Owner admits that “Fox SMART
`
`teaches that documents can be displayed in response, for example to a user query.”
`
`Resp. 41. User directed searches in Fox SMART utilized the clustering processes
`
`discussed above. See Fox SMART 53-54; see also id. 41-42 (stating that searches
`
`may “retrieve[] groups [i.e. clusters] of documents . . . [that] were formed in part
`
`due to bibliographic connections[.]”)
`
`
`
`
`
`
`
`-8-
`
`
`
`
`
`
`
`
`Fox SMART discloses the steps of claim 1 as arranged in the claim
`
`B.
`As the explanation above demonstrates, Fox SMART discloses the steps of
`
`claim 1 as arranged in the claim. To take just one example, when an oversize
`
`cluster is split, each node in the original cluster is selected in turn for analysis;
`
`candidate cluster links are generated for that node based on analysis of indirect
`
`relationships in the database; actual cluster links are derived from the candidates;
`
`one or more nodes are identified for display; and the identity of one or more nodes
`
`is displayed using the actual cluster links (i.e., clusters). Fox SMART describes an
`
`“interactive version” of the system that performs “Clustered or sequential
`
`searches” followed by “Display of portions of text from top-ranked documents.”
`
`Fox SMART 10-11. See Fox Reply Decl. ¶ 283.
`
`C.
`
`Fox SMART discloses “eliminating candidate cluster links” by
`choosing the closest links, as required by Claim 5.
`
`As explained above, and as the Board has found, Fox SMART discloses
`
`eliminating candidate cluster links by choosing the closest links, as required by
`
`claim 5. See Inst. Dec. 18-19. For example, when an oversize cluster is split,
`
`pairwise similarity values (each of which is a candidate cluster link) are calculated,
`
`and potential new clusters are formed from “highly correlated pairs”—a measure
`
`of the “closeness” of the links. Fox SMART 49-51, 53-54. Candidate links that
`
`are not “highly correlated” enough are discarded. Id.; Fox Reply Decl. ¶ 285.
`
`
`
`-9-
`
`
`
`
`
`
`
`
`Fox SMART also teaches a numerical limit on the candidate links. See, e.g.,
`
`Fox SMART 45 (“max. no. of kids before split is forced (e.g., 20)”). For example,
`
`in a splitting operation, the number of candidate links for each node will be limited
`
`by the maximum number of nodes allowed in the cluster being split. See Fox
`
`Reply Decl. ¶ 286.
`
`IV. CLAIMS 14 – 16 ARE ANTICIPATED BY FOX THESIS
`A.
`Fox Thesis explicitly discloses each element of Claim 14.
`
`
`1.
`
`Fox Thesis discloses “initializing a set of candidate cluster
`links”
`
`Similar to the discussion above for Fox SMART, Fox Thesis discloses using
`
`an algorithm that starts “with an initially empty tree” and “produces a hierarchical
`
`clustering where all N documents in a collection end up as leaves of a multi-level
`
`tree.” Fox Thesis 193; see also id.at 213, 195. As the tree grows, it is split into
`
`smaller clusters. Fox Thesis 199-200. The basis for this clustering algorithm is an
`
`“extended vector” for each document, which includes subvectors that comprise
`
`direct links (ln) as well as indirect links (bc and cc) between documents. See Fox
`
`Thesis 195, 213. These subvectors represent the set of all relationships of that type
`
`among the documents—the set of candidate cluster links. See, e.g., id. 170-71,
`
`195; Fox Reply Decl. ¶ 288. The set of candidate cluster links is initialized in the
`
`course of constructing these subvectors. Id. Once fully generated, the matrices of
`
`
`
`-10-
`
`
`
`
`
`
`
`
`subvectors represent an expanded set of initialized “candidate cluster links,” which
`
`are later reduced and selected by the clustering and searching algorithms.
`
`
`2.
`
`Fox Thesis discloses “selecting the destination node of a
`path as the selected node to analyze”
`
`Fox Thesis discloses a process of building the subvectors mentioned above,
`
`by selecting one document at a time and computing its direct and indirect
`
`relationships. See, e.g., Fox Thesis 170-71 (“For each document it is
`
`straightforward using the definitions of the last section to determine values of the
`
`linkage, bibliographic coupling, and co-citation measures between that document
`
`and any other document.”); see also Resp. 48 (admitting that “a document that has
`
`a co-citation relationship with some other document, can indeed be viewed as the
`
`destination node of a path.”). Thus, Fox Thesis discloses selecting nodes for
`
`analysis, to construct the bc and cc subvectors for that node. Those subvectors are
`
`stored for further clustering calculations. See Inst. Dec. 15.
`
`
`3.
`
`Fox Thesis discloses “retrieving the set of direct links from
`the selected node to any other node in the database”
`
`Fox Thesis discloses that, “[b]ased on the reference pattern for a set of
`
`documents” one can derive “measures of the interconnection between those
`
`documents.” Fox Thesis 166-68. Thus, Fox Thesis discloses retrieving the direct
`
`links, which are necessary to compute the indirect bc and cc subvectors. See Fox
`
`Thesis 166-72, 181.
`
`
`
`-11-
`
`
`
`
`
`
`
`
`Patent Owner manufactures a strawman issue regarding the ln subvector by
`
`claiming that Dr. Fox testified that ln was “irrelevant,” when he actually said that
`
`Patent Owner’s “discussion of ln is irrelevant.” Compare Resp. 49 with Exhibit
`
`1009 (“Fox Decl.”) ¶ 101. In fact, the ln subvector, like the bc and cc subvectors,
`
`is derived from the set of direct links between nodes. See Fox Thesis 166-72
`
`(disclosing the “Chart of Citation Arcs,” with a notation for “Direct Reference,”
`
`and disclosing the derivation of ln, cc, and bc from it).
`
`
`4.
`
`Fox Thesis discloses “determining the weight of the path
`using the retrieved direct links”
`
`As the Board found, Fox Thesis also discloses counting direct links between
`
`documents to weight the paths. See Inst. Dec. 15; Fox Thesis 166-72, 174-77. Fox
`
`Thesis discloses that the bc and cc relationships can be assigned “weights that are
`
`based on integer counts” of links. Fox Thesis 167-68. Paths between documents
`
`in the bc and cc matrices are given higher counts if there are multiple instances of
`
`coupling or co-citation between the nodes. Id. In addition, Fox Thesis discloses
`
`computing more complicated similarity coefficients (weights) utilizing additional
`
`factors that also take into account direct links between the selected node and other
`
`nodes. Fox Thesis 174-77; Fox Reply Decl. ¶ 295.
`
`
`
`
`
`
`
`-12-
`
`
`
`
`
`
`
`
`
`5.
`
`Fox Thesis discloses “repeating” steps 3 through 5 above
`“for each path”
`
`As discussed above, the process disclosed in Fox Thesis for building the bc
`
`and cc matrices is iterative, and proceeds according to the above steps, one
`
`document at a time. See Fox Thesis 170-71.
`
`
`6.
`
`Fox Thesis discloses “storing the determined weights as
`candidate cluster links”
`As the Board found, Fox Thesis discloses that the constructed bc and cc
`
`subvectors are part of the “complete extended vector for each” document that make
`
`up the “raw data for clustering.” See Inst. Dec. 15; Fox Thesis 170-71, 195, 272.
`
`Thus, the indirect relationship subvectors define a set of candidate cluster links
`
`when performing clustering using that relationship. See, e.g., Fox Thesis 213 (“bc,
`
`cc, ln, and tm subvectors should be used as the basis for clustering”); Fox Reply
`
`Decl. ¶ 297. Storing the determined weights is necessary because building the
`
`extended vectors is done prior to clustering. Id.; see also Fox SMART 2, 5, 9, 13-
`
`21, 27-40, 80-81 (describing storage of weighted subvectors and clusters).
`
`B.
`
`Fox Thesis discloses “deriving the actual cluster links” as a subset
`of the candidate links, as required by Claim 15.
`
`As discussed above, Fox Thesis discloses a clustering algorithm that
`
`“produces a hierarchical clustering where all N documents in a collection end up as
`
`leaves of a multilevel tree.” See Fox Thesis 193, 195, 199-200, 213. As the tree
`
`grows, it is split into smaller clusters. Fox Thesis 199-200. Each sub-cluster
`
`
`
`-13-
`
`
`
`
`
`
`
`
`represents a derived set of actual cluster links that are more closely related to each
`
`other than to other branches of the tree. See, e.g., Fox Thesis 211 (forming clusters
`
`“based on high pairwise similarity values”); Fox Reply Decl. ¶ 198. Thus, as the
`
`Board found, Fox Thesis discloses this limitation. Inst. Dec. 15-16.
`
`C.
`
`Fox Thesis discloses further “choosing the top rated candidate
`cluster links,” as required by Claim 16.
`Fox Thesis also discloses choosing top rated candidate cluster links. See
`
`Fox Thesis 186-87, 207, 211. For example, Fox Thesis discloses using “pairwise
`
`similarity values” and “relative pairwise distances between documents” to add
`
`documents to “the cluster that has the most categories in common.” Fox Thesis
`
`207, 211. Patent Owner ignores this disclosure in its response. Fox Thesis also
`
`discloses the use of “thresholds” applied to the “pairwise similarities” to select
`
`closer linked pairs. Fox Thesis 186-187. Thus, Fox Thesis discloses multiple
`
`methods for “choosing the top rated” links.
`
`Patent Owner objects that links are not “eliminated” (see Resp. 57), but even
`
`if that were true, the claim has no such requirement. It requires only deriving a
`
`subset of links by a process that comprises choosing the top rated links (which of
`
`course “eliminates” the unchosen links from the cluster).
`
`V. CONCLUSION
`For the foregoing reasons, the challenged claims are anticipated by Fox
`
`SMART and Fox Thesis.
`
`
`
`-14-
`
`
`
`
`
`
`
`
`
`
`DATED: September 5, 2014
`
`COOLEY LLP
`ATTN: Heidi L. Keefe
`Patent Docketing
`1299 Pennsylvania Avenue, N.W.
`Suite 700
`Washington, D.C. 20004
`Tel: 650-843-5001
`Fax: 650-849-7400
`
`
`
`
`
`Respectfully submitted,
`
`/Heidi L. Keefe/
`Heidi L. Keefe
`Reg. No. 40,673
`Counsel for Petitioners Facebook, Inc.,
`LinkedIn Corp., and Twitter Inc.
`
`
`
`
`-15-
`
`
`
`
`
`
`
`
`CERTIFICATE OF SERVICE
`
`The undersigned hereby certifies that a copy of the foregoing
`
`PETITIONERS’ REPLY TO PATENT OWNER’S RESPONSE and Exhibits 1029
`
`to 1031 were served on September 5, 2014 by electronic mail directed to the
`
`attorneys of record at the following address:
`
`Martin M. Zoltick
`Nancy J. Linck
`Rothwell, Figg, Ernst & Manbeck, P.C.
`607 14th St., N.W., Suite 800
`Washington, DC 20005
`Service Email: SRA-IPR@rfem.com
`
`
`
`
`
`Lead and Backup Counsel for Patent
`Owner
`
`
`
`
`/Heidi L. Keefe/
`Heidi L. Keefe
`Reg. No. 40,673
`Counsel for Petitioners Facebook, Inc.,
`LinkedIn Corp., and Twitter Inc.
`
`
`
`
`-16-
`
`
`
`
`