Facebook Inc.v.Software Rights Archive, LLC

Patent Trial and Appeal Board

Feb 2, 2015

08076658 (P.T.A.B. Feb. 2, 2015)

Trials@uspto.gov Paper 58
352.272.7822 Entered: February 2, 2015

UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

FACEBOOK, INC., LINKEDIN CORP., and TWITTER, INC.,
Petitioner,

v.

SOFTWARE RIGHTS ARCHIVE, LLC,
Patent Owner.
____________

Case IPR2013-00478
Patent 5,544,352
____________

Before SALLY C. MEDLEY, CHRISTOPHER L. CRUMBLEY, and
BARBARA A. PARVIS, Administrative Patent Judges.

CRUMBLEY, Administrative Patent Judge.

FINAL WRITTEN DECISION
35 U.S.C. § 318(a) and 37 C.F.R. § 42.73

I. BACKGROUND
A. Introduction
On July 30, 2013, Facebook, Inc., LinkedIn Corp., and Twitter, Inc.
(collectively “Petitioner”) filed a Petition requesting an inter partes review
of claims 26, 28–30, 32, 34, and 39 of U.S. Patent No. 5,544,352 (Ex. 1001,
“the ’352 patent”). Paper 1 (“Pet.”). On February 3, 2014, we instituted
IPR2013-00478
Patent 5,544,352

2

trial on all challenged claims, on certain of the grounds of unpatentability
alleged in the Petition. Paper 17 (“Decision to Institute” or “Inst. Dec.”).
After institution of trial, Software Rights Archive, LLC (“Patent
Owner”), filed a Patent Owner Response (“PO Resp.”). Paper 34. Petitioner
also filed a Reply. Paper 43 (“Reply”).
A consolidated oral hearing for IPR2013-00478, IPR2013-00479,
IPR2013-00480, and IPR2013-00481, each involving the same Petitioner
and the same Patent Owner, was held on October 30, 2014. The transcript of
the consolidated hearing has been entered into the record. Paper 57, “Tr.”
We have jurisdiction under 35 U.S.C. § 6(c). This Final Written
Decision is issued pursuant to 35 U.S.C. § 318(a) and 37 C.F.R. § 42.73.
Petitioner has shown by a preponderance of the evidence that claims
26, 28–30, 32, 34, and 39 of the ’352 patent are unpatentable.
B. Related Proceedings
Petitioner and Patent Owner both indicate that the ’352 patent is
involved in the following co-pending district court proceedings: Software
Rights Archive, LLC v. Facebook, Inc., Case No. 12-cv-3970; Software
Rights Archive, LLC v. LinkedIn Corp., Case No. 12-cv-3971; and Software
Rights Archive, LLC v. Twitter, Inc., Case No. 12-cv-3972, each pending in
the United States District Court for the Northern District of California. Pet.
1; Paper 8, Patent Owner’s Mandatory Notice, 2.
In addition, we instituted trial on Petitioner’s petitions on related
patents including: (1) IPR2013-00479 and IPR2013-00480, inter partes
reviews of U.S. Patent No. 5,832,494 (the “’494 patent”); and (2) IPR2013-
00481, an inter partes review of U.S. Patent No. 6,233,571 (the “’571
patent”). The ’352 patent issued from the parent of the application that
IPR2013-00478
Patent 5,544,352

3

issued as the ’494 patent. The ’571 patent issued from an application that
was a divisional of the application that issued as the ’494 patent. The ’352
patent was the subject of Reexamination No. 90/011,010.
C. The ’352 patent
The ’352 patent relates to computerized research on databases. Ex.
1001, 1:7–11. The ’352 patent discloses that it improves search methods by
indexing data using proximity indexing techniques. Id. at 3:42–55.
According to the ’352 patent, proximity indexing techniques generate a
quick-reference of the relations, patterns, and similarity found among the
data in the database. Id. at 3:53–55.
Figure 2 of the ’352 patent illustrates the high-level processing of
software for computerized searching (Id. at 8:7–8) and is reproduced below:

Figure 2 depicts software system 60 comprising Proximity Indexing
Application Program 62, Computer Search Program for Data Represented by
Matrices (“CSPDM”) 66, and Graphical User Interface (“GUI”) program 70.

IPR2013-00478
Patent 5,544,352

4

Id. at 10:53–60.
Processing of software system 60 begins with Proximity Indexing
Application Program 62 indexing a database. Id. at 11:4–5. Then, CSPDM
66 searches the indexed database and retrieves requested objects. Id. at
11:6–10. CSPDM 66 relays the retrieved objects to GUI program 70 to
display on a display. Id. at 11:10–13.
Software system 60 runs on a computer system comprising, for
example, a processor of a personal computer. Id. at 9:39–44. The system
comprises a display, which displays information to the user. Id. at 10:4–7.
Exemplary displays include computer monitors, televisions, LCDs, or LEDs.
Id.
The processor is connected to a database to be searched. Id. at 9:46–
47. The database contains cases—also called full textual objects—that
contain citations to other objects within the database. Id. at 12:1–10. Each
full textual object is assigned a number corresponding to its chronological
order in the database. Id.
The ’352 patent discloses that any two textual objects in the database
may be related through a number of “patterns.” Id. at 12:31–32. For
example, object A may cite B, or the two objects may cite the same object C.
Id. at 12:46–61. The Proximity Indexing Application (discussed above)
applies algorithms to these relationships to create a matrix of pattern vectors
that represent the relationships between the various objects in the database.
Id. at 12:62–13:3, 14:18–20. The CSPDM is used to search the indexed
database. Id. at 14:20–21.
IPR2013-00478
Patent 5,544,352

5

D. Illustrative Claim
Of the challenged claims, only claim 26 is independent,
whereas claims 28–30, 32, 34, and 39 depend, directly or indirectly,
from claim 26. Claim 26 is illustrative of the claimed subject matter
and is reproduced below:
26. A non-semantical method for numerically
representing objects in a computer database and for
computerized searching of the numerically represented objects
in the database, wherein direct and indirect relationships exist
between objects in the database, comprising:
marking objects in the database so that each marked
object may be individually identified by a computerized search;
creating a first numerical representation for each
identified object in the database based upon the object's direct
relationship with other objects in the database;
storing the first numerical representations for use in
computerized searching;
analyzing the first numerical representations for indirect
relationships existing between or among objects in the
database;
generating a second numerical representation of each
object based on the analysis of the first numerical
representation;
storing the second numerical representation for use in
computerized searching; and
searching the objects in the database using a computer
and the stored second numerical representations, wherein the
search identifies one or more of the objects in the database.
Ex. 1001, 35:28–54.

IPR2013-00478
Patent 5,544,352

6

E. The Prior Art References Upon Which Trial Was Instituted
Yahiko Kambayashi et al., Dynamic Clustering Procedures for
Bibliographic Data, Kyoto Univ., Dep’t of Inf. Sci., 90–99 (1981)
(“Kambayashi”) (Ex. 1004).
Colin F.H. Tapper, Citation Patterns in Legal Information Retrieval, 3
DATENVERARBEITUNG IM RECHT 249–75 (1976) (“Tapper 1976”) (Ex. 1005).
Colin Tapper, The Use of Citation Vectors for Legal Information
Retrieval, 1 J. OF LAW AND INFO. SCI. 131–61 (1982) (“Tapper 1982”) (Ex.
1006).
Edward A. Fox, Characterization of Two New Experimental
Collections in Computer and Information Science Containing Textual and
Bibliographic Concepts (Sept. 1983) (Ph.D. dissertation, Cornell Univ.
Dep’t of Comp. Sci.) (“Fox Collection”) (Ex. 1007).
Edward A. Fox, Some Considerations for Implementing the SMART
Information Retrieval System under UNIX (Sept. 1983) (Ph.D. dissertation,
Cornell Univ. Dep’t of Comp. Sci.) (“Fox SMART”) (Ex. 1008).
Edward A. Fox, Extending the Boolean and Vector Space Models of
Information Retrieval with P-Norm Queries and Multiple Concept Types
(Aug. 1983) (Ph.D. dissertation, Cornell Univ. Dept. of Comp. Sci.) (“Fox
Thesis”) (Ex. 1009).
The parties do not dispute the prior art status of the references.
IPR2013-00478
Patent 5,544,352

7

F. The Pending Grounds of Unpatentability
Reference(s) Basis Claims instituted
Kambayashi § 102 26, 28–30, 32, 39
Fox Thesis, Fox SMART,
and Fox Collection
§ 103

26, 28–30, 32, 34,
39
Tapper 1976 and Tapper
1982
§ 103 26, 28–30, 32, 34,
39

II. ANALYSIS
A. Claim Construction
1. Principles of Law
Petitioner asserts, and Patent Owner does not dispute, that the ’352
patent expired on August 6, 2013. Pet. 6. The Board’s interpretation of the
claims of an expired patent is similar to that of a district court’s review. See
In re Rambus, Inc., 694 F.3d 42, 46 (Fed. Cir. 2012). We, therefore, are
guided by the principle that the words of a claim “are generally given their
ordinary and customary meaning,” as understood by 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) (citation omitted).
“In determining the meaning of the disputed claim limitation, we look
principally to the intrinsic evidence of record, examining the claim language
itself, the written description, and the prosecution history, if in evidence.”
DePuy Spine, Inc. v. Medtronic Sofamor Danek, Inc., 469 F.3d 1005, 1014
(Fed. Cir. 2006) (citing Phillips, 415 F.3d at 1312–17). There is a “heavy
presumption,” however, that a claim term carries its ordinary and customary
meaning. CCS Fitness, Inc. v. Brunswick Corp., 288 F.3d 1359, 1366 (Fed.
Cir. 2002) (citation omitted).
IPR2013-00478
Patent 5,544,352

8

2. Overview of the Parties’ Positions
In the Decision to Institute, we found it instructive to construe the
claim terms direct relationships, indirect relationships, pool-similarity
searching, and pool-importance searching. Inst. Dec. 10–14. Our
constructions are set forth in the table below.
Claim Term or Phrase Construction
direct relationships “relationships where one object cites to
another object” Inst. Dec. 13.
indirect relationships “relationships where at least one
intermediate object exists between two
objects and where the intermediate
object(s) connect the two objects through
a chain of citations” Inst. Dec. 13.
pool-similarity searching “identifying at least one object based on
degree of similarity to a selected pool of
objects” Inst. Dec. 14.
pool-importance searching “identifying at least one object based on
the importance of the object to a selected
pool of objects” Inst. Dec. 14.

Petitioner does not challenge any of our constructions. Reply 1–2.
Patent Owner appears to agree with many of our constructions, and states
that it uses our constructions for the purpose of evaluating patentability of
the challenged claims of the ’352 patent. PO Resp. 12–14. Based on the
complete record now before us, we discern no reason to change our prior
constructions.
Additionally, Patent Owner addresses the following phrases or terms:
1) objects in a computer database; 2) computerized searching; 3) non-
semantical method; 4) some indirect relationships are weighed more heavily
than other indirect relationships; and 5) relationships exist between or
among subsets of objects, which are discussed below. Id. at 9–14.
IPR2013-00478
Patent 5,544,352

9

Petitioner’s Reply further addresses database and numerical representation,
but otherwise does not contest Patent Owner’s proposed constructions of
these terms. Reply 2.
3. numerical representation
Patent Owner’s Response does not proffer an explicit construction of
numerical representation, but appears to interpret the term to exclude strings
that may include letters. PO Resp. 21 (distinguishing prior art as having
“non-numerical character strings). At oral argument, Patent Owner
confirmed that its construction of numerical representation is something
“represented only by digits,” or in other words “expressed by numbers, not
by letters.” Tr. 85.
Petitioner responds that numerical includes “any representation of
binary or digital data that can be processed and analyzed by a computer,”
and means simply “of or relating to numbers.” Reply 1; Tr. 13. Petitioner’s
construction is, therefore, not limited to representations consisting only of
numbers. At oral argument, Petitioner argued that the inclusion of a single
number into a string is sufficient to make that string a numerical
representation. Tr. 25.
Petitioner’s proffered construction is overly broad and unsupported by
the specification. While one dictionary definition of numerical is “of or
relating to a number or series of numbers,” it may also refer to “expressed in
or counted by numbers.” THE AMERICAN HERITAGE DICTIONARY OF THE
ENGLISH LANGUAGE (2000) (Ex. 3001); see also COLLINS ENGLISH
DICTIONARY (2000) (Ex. 3002) (“measured or expressed in numbers”).
The specification of the ’352 patent uses numerical consistent with
this latter interpretation. In the Initial Extractor Subroutine, the “full textual
IPR2013-00478
Patent 5,544,352

10

objects” of the database are numbered “with Arabic numbers from 1 through
n.” Ex. 1001, 14:49–50. These numbers are used to create vectors and
matrices, which are then run through various algorithms such as the Opinion
Patterner Subroutine. Id. at 14:55–15:22. “Numerical factors” are then
“calculated” to determine “values.” Id. at 15:19–22; 18:63–67. This
emphasis on calculation, values, and on processing by computer algorithms,
leads us to conclude that numerical representation, as used in the ’352
specification, must refer to solely numbers, so that a computer can process
the representations using mathematical algorithms.
Petitioner’s attempt to link the numerical representation of the
specification to the West “key number” system is unpersuasive. Reply 2.
While the specification of the ’352 patent does discuss the key number
system, and such “key numbers” include letters, there is no indication that
the patentee intended to link the numerical representation of the claims to
the West key number system discussed—and distinguished—in the
background portion of the specification. Ex. 1001, 2:38–43 (“such a
numbering process is subjective and is prone to error”).
Nor do we find persuasive Petitioner’s argument that numerical is
somehow distinct from “numeric,” in that the latter term means only
numbers but the former may encompass letters. Tr. 13. Not only was this
argument advanced for the first time at oral hearing,
1
but it is unsupported
by any evidence of record. Indeed, the two terms are used interchangeably
in dictionary definitions. See Ex. 3001 (entry for “numerical also numeric”);
Ex. 3002 (entry for “numerical or numeric”).

1
Our Rules do not permit arguments to be raised for the first time at oral
hearing. 37 C.F.R. § 42.70(a) (permitting oral argument only on “an issue
raised in a paper.”).
IPR2013-00478
Patent 5,544,352

11

For these reasons, we construe numerical representation as
“representation consisting exclusively of numbers or a set of numbers.”
4. objects in a computer database and computerized searching
Patent Owner addresses the terms objects in a computer database and
computerized searching, both of which appear in claim 26. PO Resp. 9–11.
Rather than proffer a construction for either term, however, Patent Owner
discusses the general concept of computerized searching in the ’352 patent.
Id. It is not clear what construction Patent Owner wishes us to adopt, and
we are not persuaded that either term requires an explicit construction.
5. non-semantical method
Patent Owner asks that we interpret non-semantical method to mean
“a method that uses the direct relationships between one database object and
another and does not otherwise account for words and phrases in a textual
object.” PO Resp. 11. We note that Petitioner raised a similar construction
in the Petition, but the Board declined to construe the term expressly in the
Decision to Institute. Pet. 6–7; Inst. Dec. 11. We also note that we adopted
a similar construction for “non-semantically” in the related case IPR2013-
00481. We consider the proffered construction to be reasonable and
consistent with the specification of the ’352 patent, and adopt it herein.
6. some indirect relationships are weighed more heavily than
other indirect relationships
Patent Owner asks that we construe this phrase as “some types of
indirect relationships are weighed more heavily than others.” PO Resp. 13
(emphasis in original). To support this interpretation, Patent Owner cites the
specification of the ’352 patent, which discloses that the different “patterns,”
which include direct and indirect relationships, are assigned various weights.
IPR2013-00478
Patent 5,544,352

12

Ex. 1001, 13:34–38. Petitioner does not dispute this construction. We
consider the proffered construction to be reasonable and consistent with the
specification of the ’352 patent, and adopt it herein.
7. relationships exist between or among subsets
Patent Owner does not set forth an express construction for this phrase
which appears in claim 34, but instead states that the “relationships” are the
direct and indirect relationships of claim 26, and that subsets are a portion of
a textual object. PO Resp. 13–14. Patent Owner has not persuaded us that
an express construction of this phrase is necessary.
B. Anticipation of Claims 26, 28–30, 32, and 39 by Kambayashi
We instituted trial to determine whether claims 26, 28–30, 32, and 39
are unpatentable under 35 U.S.C. § 102 as anticipated by Kambayashi. Dec.
Inst. 19–20. To establish anticipation, Petitioner must prove that each and
every element in a claim, arranged as is recited in the claim, may be found in
a single prior art reference. Net MoneyIN, Inc. v. VeriSign, Inc., 545 F.3d
1359, 1369 (Fed. Cir. 2008); Karsten Mfg. Corp. v. Cleveland Golf Co., 242
F.3d 1376, 1383 (Fed. Cir. 2001). We determine that Petitioner has not
shown by a preponderance of the evidence that claims 26, 28–30, 32, and 39
are unpatentable as anticipated by Kambayashi.
1. Kambayashi
Kambayashi describes a method for clustering, which is said to be “an
important tool for efficient retrieval of documents in bibliographic database
systems.” Ex. 1004, Abstract. The reference discloses the creation of
“Direct Reference Matrix R,” defining direct reference as “when a paper A
refers to a paper B.” Id. at 91–92. Kambayashi also discloses a set of pairs
IPR2013-00478
Patent 5,544,352

13

(ID, IDF) where ID and IDF are the identification codes of the papers and
(ID, IDF) means that paper ID cites paper IDF. Id. at 93.
Kambayashi also discloses the creation of two secondary matrices,
“Bibliographic Coupling Matrix B” (papers with one or more citation in
common) and “Co-citation Matrix C” (citations frequently cited together).
Id. at 92. These secondary matrices consist of vectors derived from the (ID,
IDF) pairs noted above. Id. at 93–34. A “Similarity Matrix S” may then be
created via weighted summation of matrices R, B, and C. Id. at 92.
2. Claim 26
We focus our analysis herein on two steps required by the method of
claim 26: “creating a first numerical representation for each identified object
in the database based upon the object’s direct relationship with other objects
in the database,” and “analyzing the first numerical representations for
indirect relationships existing between or among objects in the database.”
Petitioner contends that Kambayashi discloses both of these steps in two
alternative embodiments.
First, Petitioner directs us to Kambayashi’s disclosure of (ID, IDF)
pairs, and their use in creating the B and C matrices. Petitioner asserts that
deriving the (ID, IDF) pairs may be considered to be creating a first
numerical representation, and that they represent direct relationships
between documents ID and IDF. Pet. 24–25; Tr. 22. The (ID, IDF) pairs are
then analyzed for indirect relationships, leading to B and C matrices which
Petitioner contends are second numerical representations. Pet. 25–26.
Second, Petitioner identifies Kambayashi’s Direct Reference Matrix R
as a first numerical representation that represents direct relationships. Pet.
24–25; Tr. 22. As noted above, Matrix R is used—along with matrices B
IPR2013-00478
Patent 5,544,352

14

and C—to generate a Similarity Matrix S. Petitioner contends that if Matrix
R is considered to be the first numerical representation, then Matrix S would
be a second numerical representation within the scope of claim 26. Tr. 22.
Upon review of the disclosure of Kambayashi, we find neither of
these arguments persuasive. First, in the case of the (ID, IDF) pairs that are
used to derive the B and C matrices, Patent Owner argues that ID and IDF
are strings that contain letters. PO Resp. 21. This is supported by the
disclosure of Kambayashi, which discloses identification codes such as
EVER7404 and GARDL7710. Ex. 1004, 96. The testimony of Dr. Jacobs,
Patent Owner’s declarant, explains how Kambayashi’s source database
shows that identification codes begin with the first four letters of the first
author’s name. Ex. 2113 ¶¶ 100–103 (citing Ex. 2023). Because, as
discussed above, the proper construction of numerical representation is a
representation that contains only numbers, Kambayashi’s (ID, IDF) pairs
cannot be the first numerical representation of claim 26.
Nor do we find that the Direct Reference Matrix R / Similarity Matrix
S system of Kambayashi meets the limitations of claim 26. Petitioners
assert, and we agree, that Matrix R is an array of numbers that represents
direct relationships between the objects in the Kambayashi database. Pet.
24–25; Ex. 1004, 92 (values rij of matrix R are either 0 or 1). This
conclusion is supported by the fact that, in order to derive Matrix S, Matrix
R is multiplied by a constant wR. Ex. 1004, 92. It would not make sense to
multiply a matrix containing strings of letters by a constant.
Matrix S, however, cannot be the second numerical representation of
claim 26. The claim requires that the representation be created by
“analyzing the first numerical representations for indirect relationships
IPR2013-00478
Patent 5,544,352

15

existing between or among objects in the database.” Matrix S, however, is
generated according to the following formula:
S = wR*R+wB*B'+wC*C' (Id.)
Patent Owner argues that multiplying matrix R by a constant is not
analyzing, as that term is used in claim 26. PO Resp. 21. We agree. While
Matrix S does take into account indirect relationships between objects, those
relationships are not derived from an analysis of Matrix R (the first
numerical relationship). Rather, the indirect relationships are accounted for
in Matrix S by the inclusion of Matrices B (which tracks bibliographic
coupling) and C (which tracks co-citation). Id. The indirect relationships
reflected in the B and C matrices, in turn, are not derived from Matrix R, but
rather from the (ID, IDF) pairs, which we have determined above cannot be
the first numerical relationship. For these reasons, Matrix S of Kambayashi
does not meet the second numerical representation limitation of claim 26,
because it is not generated by analyzing the first numerical representation.
We, therefore, conclude that Kambayashi fails to disclose an
embodiment having all elements of claim 26, as arranged in the claim.
Kambayashi does not anticipate claim 26.
3. Dependent Claims
The remaining instituted claims all depend, directly or indirectly, from
claim 26, and incorporate claim 26’s requirements of a first numerical
representation, and second numerical representation. We, therefore, find
that Kambayashi does not anticipate the dependent claims, for the same
reasons discussed above with respect to claim 26.
IPR2013-00478
Patent 5,544,352

16

C. Obviousness of Claims 26, 28–30, 32, 34, and 39 Over Fox Papers
We instituted trial to determine whether claims 26, 28–30, 32, 34, and
39 are unpatentable under 35 U.S.C. § 103 as having been obvious over the
combined disclosures of Fox Thesis, Fox SMART, and Fox Collection
(collectively, “the Fox Papers”). Inst. Dec. 14–19. In support of the
asserted ground of unpatentability, Petitioner sets forth the teachings of the
cited prior art, provides detailed claim charts, and cites to the declaration of
Dr. Fox (Ex. 1016 ¶¶ 68–145), explaining how each limitation is taught in
the cited prior art combination. Pet. 9–23.
The claim chart persuasively reads all elements of each of claims 26,
28–30, 32, 34, and 39 onto the teachings of the Fox Papers, taken together.
Despite the counter-arguments in Patent Owner’s Response, and the
evidence cited therein, which we have also considered, Petitioner has shown
by a preponderance of the evidence that each of claims 26, 28–30, 32, 34,
and 39 of the ’352 patent are unpatentable, under 35 U.S.C. § 103, as they
would have been obvious over the combination of Fox Thesis, Fox
Collection, and Fox SMART.
1. Fox Thesis
Fox Thesis describes improving query and document representation
schemes for information retrieval. Ex. 1009, 261. In particular, useful types
of bibliographic data are incorporated into a model to test clustering and
retrieval functions. Id. at 164. Bibliographic connections between articles
are illustrated for an exemplary set “O” of documents, which are represented
by letters A through G. Id. at 165–66, Fig. 6.2. This exemplary set “O”
includes direct and indirect citation references. Id. at 166–67, Table 6.2.
IPR2013-00478
Patent 5,544,352

17

Based on the reference pattern for a set of documents, Fox Thesis
describes deriving various measures of the interconnection between the
documents. Id. at 166. For example, weights are assigned “based upon
integer counts” for bibliographically coupled documents. Id. at 167.
Citation submatrices represent reference or citation information. Id. at 169.
For example, submatrix bc represents bibliographically coupled reference
information and submatrix cc represents co-citation reference information.
Id. at 169–72, Figs. 6.3–6.5.
2. Fox SMART
Fox SMART describes the System for Mechanical Analysis and
Retrieval of Text (“SMART”) as a project for designing a fully automatic
document retrieval system and for testing new ideas in information science.
Ex. 1008, 3. Fox SMART describes the computer system used to implement
the experiments described in the Fox Thesis. Ex. 1016, ¶ 27. The software
components of SMART are implemented in the C Programming Language
and run under the UNIX™ operating system on a VAX™ 11/780 computer.
Ex. 1008, 1, 4.
In SMART, an automatic indexing component constructs stored
representations of documents. Id. at 3. Bibliographic information is used to
enhance document representations. Id. at 29. The SMART system may
process basic raw data, such as an exemplary N collection of articles and
citation data describing which articles are cited by others. Id. at 29–30.
Data is entered into the SMART system as a set of tuples {(di, dj)|di→dj}
which describe the cited and citing documents, as well as the direction of
citation. Id. at 29. The exemplary input data also includes indirect citation
relationships, such as bibliographic coupled and co-citation relationships.
IPR2013-00478
Patent 5,544,352

18

Id. at 30–32. These relationships are used to create extended vectors which
can then be clustered and searched to aid document retrieval. Id. at 29.
3. Fox Collection
Fox Collection describes collections of data which are said to be
useful for investigating the interaction of textual and bibliographic data in
retrieval of documents. Ex. 1007, 1. According to the testimony of Dr.
Edward Fox, Fox Collection was originally part of the same work as Fox
Thesis and Fox SMART, and describes the manner in which the data sets
were obtained and processed prior to their use in the Fox SMART
experiments. Ex. 1016 ¶ 27.
According to Fox Collection, the experiments were performed on a
collection of bibliographic records (title, abstract, author, keywords, etc.)
from the Communications of the ACM, termed the “CACM collection.” Ex.
1007, 14.
2
Two individuals then examined printed copies of the articles
referenced by the CACM bibliographic records, and citation data was
obtained from the articles and entered into a set Raw_data. Id. The citation
data contained pairs of identifiers (citing, cited) which were the document id
numbers (“dids”) of the citing record and record it cites. Id. From this
Raw_data matrix, secondary matrices such as bc (bibliographic coupling)
and cc (co-citation) were derived computationally. Id. at 14–16.

2
Fox Collection also discusses an ISI Collection, but in his Reply
Declaration Dr. Fox explains that he cites the ISI collection to “emphasize
findings in the prior art about the value of using co-citation data (a non-
semantic indirect relationship) in information retrieval, not to fully address
all the elements of claims. . . . For the sake of simplicity, the Board should
focus on the methodology given in Fox Papers, and the examples of their use
with the CACM Collection.” Ex. 1030, ¶ 6.
IPR2013-00478
Patent 5,544,352

19

4. Claim 26
Petitioner’s claim chart persuasively reads all elements of claim 26
onto the combined teachings of Fox Thesis, Fox SMART, and Fox
Collection. Pet. 9–23 (citing Ex. 1007, 14–15, 43, 48; Ex. 1008, 3, 12–13,
16, 18, 25–27, 29–33, 36, 38–39, 41–43, 53; Ex. 1009, 17, 19, 179, 181–82,
195, 199, 203, 211; 1016 ¶¶ 71–108, 122–131). For instance, the
combination of Fox Thesis, Fox SMART, and Fox Collection teaches
“marking objects in the database” and “creating a first numerical
representation for each identified object in the database based upon the
object’s direct relationship with other objects in the database,” as recited in
claim 26. In particular, Fox Collection teaches assigning document
identification numbers (“dids”) to the articles in the CACM collection,
which is “marking objects in the database.” Ex. 1007, 14. Printed copies of
each article with a bibliographic entry in the CACM collection then are
reviewed manually, to obtain bibliographic subvectors in the form
“Raw_data (cited, citing).” Id. This is a first numerical representation
created based on the direct relationship between the “cited, citing” pair of
bibliographic records.
The combination of Fox Thesis, Fox SMART, and Fox Collection
also teaches “analyzing the first numerical representations for indirect
relationships existing between or among objects in the database” and
“generating a second numerical representation of each object based on the
analysis of the first numerical representation,” as recited in claim 26. Fox
SMART teaches that direct relationships may be represented by tuples called
“CITED,” which contain a citing document, a cited document, and the
direction of the citation. Ex. 1008, 29. These tuples are then processed to
IPR2013-00478
Patent 5,544,352

20

construct submatrices such as bc and cc, which contain numbers
representing indirect relationships. Id. at 30–32 (“construct BC by counting
the number of identical tuples of C”). Dr. Fox testifies that the CITED
tuples of Fox SMART refer to the Raw_data derived from the CACM
collection. Ex. 1016 ¶ 124. Because these bc and cc submatrices are
numerical representations, and are generated from the first numerical
representations CITED which are based on direct relationships, we find that
the Fox Papers together teach “generating a second numerical representation
of each object based on the analysis of the first numerical representation.”
a. Combination of References
As to whether Petitioner has satisfied the requirements for combining
the teachings of Fox Thesis, Fox SMART, and Fox Collection, we determine
that Petitioner has articulated sufficient reasoning with a rational
underpinning as to why one of ordinary skill in the art would have combined
the retrieval systems taught in Fox Thesis, Fox SMART, and Fox Collection.
See KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 418 (2007) (citing In re
Kahn, 441 F.3d 977, 988 (Fed. Cir. 2006)). Dr. Fox states that the three
publications arose from the same thesis project, and were originally one
document. Ex. 1016 ¶ 70. Furthermore, Dr. Fox notes that Fox Thesis
“explain[s] the method and experimental results of [his] information
retrieval work,” Fox SMART “detail[s] the updated SMART computer
system used to execute the experiments,” and Fox Collection “describes how
the data sets were obtained and processed prior to being used in the
experiments.” Id. We give Dr. Fox’s statement that one of skill in the art
would have been motivated to combine the references because they
“describe a complete project with its underlying system and data” (id.)
IPR2013-00478
Patent 5,544,352

21

substantial weight, because it is consistent with the considerable overlap in
the disclosures of the Fox Papers and their internal references to one another.
See, e.g., Ex. 1009, 343 (Fox Thesis cites to Fox SMART); Ex. 1008, 84
(Fox SMART cites to Fox Thesis).
b. Patent Owner’s Counterarguments
We have considered Patent Owner’s counterarguments but do not find
them persuasive. Patent Owner contends that various elements of claim 26
are not taught or suggested by the Fox Papers in combination. First, Patent
Owner argues that the Fox Papers do not teach a computer database in which
direct and indirect relationships exist between objects in the database. PO
Resp. 17. Because the databases of the Fox Papers are bibliographic
databases, they contain certain information about documents such as title,
abstract, author, and publication date. See, e.g., Ex. 1007, 14 (describing
CACM database). The Fox databases do not contain the full documents,
meaning that the databases do not contain the portions of the documents that
cite to other documents. As such, Patent Owner argues, the databases cannot
have objects having direct and indirect relationships, as required by claim
26. PO Resp. 17–18.
We have construed direct relationships to mean “relationships where
one object cites to another object.” Based on this construction, the
bibliographic records of Fox Collection’s CACM database do not have
direct relationships, because they do not contain cites to one another. It is
only after the full documents—which are not in the database—are manually
reviewed, and the first numerical representation (Raw_data) is entered, that
the database contains objects that have direct relationships. Claim 26,
however, requires that direct and indirect relationships exist between objects
IPR2013-00478
Patent 5,544,352

22

in the database first, prior to creating a first numerical representation. In
other words, the Raw_data of the Fox Collection CACM database cannot be
both the objects that have relationships, as well as the first numerical
representation of those relationships.
Petitioner contends, however, that it would have been “trivial and
obvious” to modify the databases of the Fox Papers to contain full text
documents. Reply 10. Dr. Fox’s testimony supports this argument, noting
that if storage resources allowed storage of the full text of documents, this
would have been understood as preferable. Ex. 1016 ¶¶ 76, 89. We credit
Dr. Fox’s testimony on this point, as it is consistent with the disclosure of
Fox Thesis that “some [information retrieval] systems store the full text of
the various documents.” Ex. 1009, 6. Fox Thesis adds that full text permits
users to “locate documents of interest,” as well as “retrieve and/or examine
paragraphs, passages, sentences, or single word occurrences (in context).”
Id. These extra capabilities are described as “straightforward generalizations
of document retrieval methods.” Id.
We, therefore, conclude that the Fox Papers suggested to one of
ordinary skill in the art at the time of the invention the modification of the
Fox databases to include full text documents. With such a modification, the
databases would contain, as objects, the full text documents. Therefore,
even prior to generation of the Raw_data, the database would contain objects
that have direct and indirect relationships due to their citation of one another.
Patent Owner’s argument to the contrary is unpersuasive.
In the same vein, Patent Owner argues that the Fox Papers do not
teach “creating a first numerical representation for each identified object in
the database based upon the object’s direct relationship with other objects in
IPR2013-00478
Patent 5,544,352

23

the database.” PO Resp. 31. Because, for example, the Raw_data disclosed
in the Fox Collection is derived from documents that are not in the CACM
database, but rather compiled from full text printed versions of the
documents, Patent Owner argues that Raw_data is not based on the object’s
direct relationship with other objects. Id. at 32.
We find this argument unpersuasive for the same reasons outlined
above for the objects limitation. The Fox Papers suggest inclusion of full
text documents in the databases. With such a modification, the databases
would have—even prior to creation of Raw_data—objects with direct
relationships. The subsequently-created Raw_data relation would be based
on those objects, thus satisfying the first numerical representation element of
claim 26.
Patent Owner’s remaining contentions relate to whether the Petitioner
has satisfied the requirements for combining the teachings of Fox Thesis,
Fox SMART, and Fox Collection. For example, Patent Owner contends that
the systems disclosed in the individual Fox Papers are “narrowly tailored”
and would not have been combined merely because of their common
authorship. PO Resp. 26.
As indicated above, we determine that Petitioner has articulated
sufficient reasoning with a rational underpinning as to why one of ordinary
skill in the art would have combined the retrieval systems taught in Fox
Thesis, Fox SMART, and Fox Collection. See KSR, 550 U.S. at 398. For
instance, Dr. Fox wrote each of Fox Thesis, Fox SMART, and Fox
Collection. See Ex. 1009, i; Ex. 1008, 1; Ex. 1007, 1.
Patent Owner also contends that the Raw_data relation of Fox
Collection could not be combined with the CITED tuples of Fox SMART,
IPR2013-00478
Patent 5,544,352

24

because they are “fundamentally incompatible.” PO Resp. 27. In support of
this argument, Dr. Jacobs testifies, for example, that CITED does not
describe using document ids (“dids”) while Raw_data does. Ex. 2113 ¶¶
170–171. Dr. Fox testifies to the contrary, stating that the CITED tuples of
Fox SMART specifically refer to the Raw_data derived from the CACM
collection. Ex. 1016 ¶ 124. We give Dr. Fox’s testimony on this point
substantial weight. Our determination is not only due to Dr. Fox’s personal
knowledge of the Fox Papers, but also supported by the descriptions of
Raw_data and CITED in the references. The references indicate that both
Raw_data and CITED contain pairs of document identifiers, with the sole
difference being that CITED also contains a third data element that signifies
the direction of the citation. Furthermore, while the description of CITED in
Fox SMART is silent as to document ids, other portions of the document
discuss dids which are an “index in range 1 . . . N.” Ex. 1008, 36. We do
not consider the combination of Raw_data with CITED, or the combination
of the systems of Fox Collection, Fox SMART, and Fox Thesis, to be
beyond the level of ordinary skill in the art.
Patent Owner further contends that using indirect relationships in a
computerized search system would not have been predictable at the time of
the invention of the ’352 patent. PO Resp. 50. Patent Owner’s contention is
based on its view that the combined teachings of Fox Thesis, Fox SMART,
and Fox Collection are not sufficient because they do not teach
computerized searching of an electronic database. PO Resp. 54; see also Tr.
49 (“[T]he Fox papers by themselves don’t get you there . . . every one . . . is
directed to printed articles, not an electronic database.”). According to
Patent Owner, the prior art cited by Petitioner teaches experiments that are
IPR2013-00478
Patent 5,544,352

25

not directed to a computer database, “but rather are directed toward limited
experimentation with bibliographic relationships existing among paper
documents.” PO Resp. 1.
We disagree with Patent Owner. For example, Fox SMART teaches
an implementation in which software components of SMART are
implemented in the C Programming Language and run under the UNIX™
operating system on a VAX™ 11/780 computer. Ex. 1008, 1, 4. In
SMART, an automatic indexing component constructs stored representations
of documents. Id. at 3. In light of the various teachings of Fox Thesis, Fox
SMART, and Fox Collection discussed herein, we determine that Fox
Thesis, Fox SMART, and Fox Collection, taken together, teach or suggest
computerized searching of an electronic database.
Patent Owner also contends that the inclusion of indirect relationships
into search “degrades results,” and therefore provides a teaching away from
the invention. PO Resp. 50. As Patent Owner acknowledges, its evidence of
degraded results does not teach away from the combination of the Fox
Papers, but rather from the modification of the teachings of the Fox Papers
to incorporate “an electronic database that has references to the objects in
the database.” Tr. 49–50. We found above, however, that the Fox Papers
teach this feature. In addition, to the extent modification of the Fox Papers
is necessary to meet claim 26, we have found that modification is expressly
suggested by the Fox Papers themselves. The record is insufficient to
establish a teaching away.
Patent Owner also asserts objective indicia of non-obviousness,
focusing on Google’s search engine using its PageRank algorithm. PO
Resp. 56–60. As an initial matter, Patent Owner’s contentions again appear
IPR2013-00478
Patent 5,544,352

26

to be based on its view that the combined teachings of Fox Thesis, Fox
SMART, and Fox Collection are not sufficient because they do not teach
computerized searching of an electronic database. Id. at 58 (“Link analysis
technology applied to the Web, as claimed in the ’352 patent and embodied
in PageRank, satisfied a long felt need for improved computerized search.”
(citation omitted)); Tr. 60–61 (“[I]t certainly wouldn’t have been obvious to
one of ordinary skill based on Fox’s work to extend these ideas from this
paper collection to electronic databases.”). For the reasons discussed above,
we disagree with Patent Owner’s view and determine that Fox Thesis, Fox
SMART, and Fox Collection, taken together, teach or suggest computerized
searching of an electronic database.
Furthermore, we note that Patent Owner has not shown that the
asserted success of a commercial embodiment of the ’352 patent actually
resulted from features recited in the claims of the ’352 patent. Patent Owner
has not provided sufficient evidence to support a nexus between claim 26
and the Google PageRank algorithm. Because Patent Owner has failed to
provide the source code of PageRank, or any other detailed information
beyond publicly-available, generalized hearsay statements about Google’s
search (Ex. 2050), the record is insufficient to prove that PageRank uses the
method of claim 26.
Even if PageRank’s algorithm incorporates the method of claim 26,
we cannot determine that Google’s success is due to the method of claim 26,
as opposed to other elements of the algorithm. Patent Owner’s declarant Dr.
Amy N. Langville conceded that the Google search technology involves a
combination of link analysis (non-semantic) and semantic searching,
whereas claim 26 recites a non-semantical method. Ex. 1034, 76:19–21.
IPR2013-00478
Patent 5,544,352

27

Even if we were to conclude that the PageRank algorithm utilized the non-
semantical method of claim 26, we could not determine whether the alleged
success of PageRank is due to its non-semantical aspects, its semantical
aspects, or some combination of both.
Patent Owner also points to Google’s license of the ’352 patent as
evidence of nexus. PO Resp. 59–60. Patent Owner, however, admits that
this license resulted in the settlement of a lawsuit (id.), which without
additional contextual evidence, weighs against finding a nexus.
Additionally, we determine that in light of the weak showing of
secondary considerations, the evidence of obviousness with respect to Fox
Thesis, Fox SMART, and Fox Collection, is sufficient to support the
conclusion that claim 26 would have been obvious. See Leapfrog
Enterprises, Inc. v. Fisher-Price, Inc., 485 F.3d 1157, 1162 (Fed. Cir. 2007).
As discussed above, Petitioner has provided a strong case of obviousness.
Accordingly, even after considering the counter-arguments in Patent
Owner’s Response, and the evidence cited therein, we find that Petitioner
has shown by a preponderance of the evidence that claim 26 is unpatentable
as it would have been obvious over the combination of Fox Thesis, Fox
SMART, and Fox Collection.
5. Dependent Claims
Petitioner’s claim chart persuasively reads all elements of dependent
claims 28, 29, 30, 32, 34, and 39 onto the teachings of Fox Thesis, Fox
SMART, and Fox Collection, taken together. Pet. 16–23 (citing Ex. 1007,
12, 14–15, 43, 48–49; Ex. 1008, 14, 16, 24–25, 29, 30–33, 36–38, 41, 43–
52; Ex. 1009, 1, 15, 17, 23, 126, 151, 173–74, 177–79, 181–82, 191–92,
194, 202, 213–18, 224, 232, 234, 238–43, 257; Ex. 1016 ¶¶ 71–108, 110–
IPR2013-00478
Patent 5,544,352

28

121, 124–125, 127–128, 132–135, 137–145). For instance, we determine
that Petitioner has shown by a preponderance of the evidence that the
combination of Fox Thesis, Fox SMART, and Fox Collection teaches that
the first and second numerical representations are vectors that are arranged
in first and second matrices, as required by claim 28. Fox SMART teaches
CITED, which is a set of tuples indicating a pair of documents linked by a
direct relationship, as well as the direction of citation. Ex. 1008, 29. These
tuples are vectors, as are the components of the bc and cc submatrices,
which represent indirect relationships. Id. at 30–32. Furthermore, the
objects in the CACM database are bibliographic records which include
publication date (Ex. 1007, 14), and therefore are assigned chronological
data as required by claim 28.
We also determine that Petitioner has shown by a preponderance of
the evidence that the combination of Fox Thesis, Fox SMART, and Fox
Collection teaches weighing, wherein some indirect relationships are
weighed more heavily than other indirect relationships, as recited in claim
32. Fox Thesis, for example, discloses assigning weights to subvectors such
as bc and cc, which are different types of indirect relationships. Ex. 1009,
257 (Table 8.13). In one weighting scheme disclosed in Fox Thesis,
bibliographic coupling indirect relationships (bc) are weighted at .009, more
heavily than co-citation indirect relationships (cc), weighted at 0. Id.
Claim 39 requires both pool-similarity searching and pool-importance
searching. As noted above, we construed pool-similarity searching as
“identifying at least one object based on degree of similarity to a selected
pool of objects” and pool-importance searching as “identifying at least one
object based on the importance of the object to a selected pool of objects.”
IPR2013-00478
Patent 5,544,352

29

We determine that the Fox Papers have been shown by a preponderance of
the evidence to teach these elements. Fox Thesis and Fox SMART disclose
a feedback search in which results are presented to a user, ranked according
to importance, and then used to construct a new search. Ex. 1008, 24, Fig. 6;
Ex. 1009, 151, Fig. 5.1. This teaches pool-importance searching as required
by claim 39. Similarly, Fox SMART discloses a search that “perform[s]
exact matches as well as general similarity computations” (Ex. 1008, 37–
38), which meets the pool-similarity searching limitation of claim 39.
Additionally, for the reasons discussed above with respect to claim 26,
we determine that Petitioner has satisfied the requirements for combining the
teachings of Fox Thesis, Fox SMART, and Fox Collection.
Patent Owner argues that Petitioner has not proven by a
preponderance of the evidence that all elements of the dependent claims are
taught or suggested by the Fox Papers. PO Resp. 35–39. Some of these
arguments, for example those made with respect to claims 28, 29, 30, and
34, are based on the fact that the databases of the Fox Papers do not include
objects because the bibliographic records do not cite to one another. Id. Just
as we found such arguments unpersuasive with respect to claim 26, we are
not persuaded by them here. The Fox Papers suggest the inclusion of full
text documents into the databases, and that such a modification could be
beneficial.
Patent Owner also argues that claim 28’s limitation that the step of
searching comprises the steps of matrix searching of the second matrices and
examining the chronological data is not met by the Fox Papers. Id. at 35–36.
According to Patent Owner, the Fox Thesis discloses a “preliminary
clustering experiment” in which chronological data is “summarily dismissed
IPR2013-00478
Patent 5,544,352

30

because of poor results.” Id. We do not consider this reading of the Fox
Thesis accurate. The quotation from the reference provided by Patent
Owner, “the clustering result does not seem as good as that of the other
methods” (Ex. 1009, 217), is partial and misleading. The full sentence
reads: “If the clustering result does not seem as good as that of other
methods then a likely explanation is that improper coefficients were chosen
and used in computing the combined similarity value.” Id. (omitted portion
emphasized). Not only does Patent Owner omit the qualifier “if,” but also
the explanation that the result likely is due to improper weighting
coefficients. This is far from the “summary dismissal” of chronological data
asserted by Patent Owner.
Indeed, as Petitioner notes, other portions of the Fox Papers expressly
disclose searching using indirect relationship matrices in combination with
chronological data. Ex. 1008, 41 (p-norm queries include date, as well as
bibliographically coupled or co-cited articles). We are not persuaded by
Patent Owner’s arguments regarding claim 28.
Patent Owner also contends that the Fox Papers do not teach or
suggest marking subsets of objects in the database, as required by claim 34.
PO Resp. 38. Fox SMART discloses “separate indexing of paragraphs or
even sentences.” Ex. 1008, 25; id. at 80 (“vectors could be computed for
smaller items than just documents”). According to Patent Owner, however,
the markings “must be usable by a computerized search to individually
identify a specific subset of an object in a computer database as a search
result.” PO Resp. 38. This alleged requirement is drawn from the marking
limitation of claim 26. Id. Claim 26, however, only requires that the
“marked object . . . be individually identified by a computerized search.”
IPR2013-00478
Patent 5,544,352

31

The subsets of claim 34 are marked, but this marking does not transform the
subsets into objects as recited in claim 26. Patent Owner’s argument that
computerized searching of the marked subsets is required by claim 34 lacks
merit.
Finally, Patent Owner argues that the Fox Papers do not teach or
suggest pool-importance searching, as required by claim 39. PO Resp. 39.
Patent Owner correctly notes that “importance is distinct from similarity,”
and therefore pool-importance searching is different than pool-similarity
searching. Id. The Petition, Patent Owner argues, only identifies disclosures
of pool-similarity searching in the Fox Papers, and, therefore, fails to
establish that all elements are taught or suggested by the prior art. Id.
We disagree with Patent Owner’s argument. In its claim chart,
Petitioner set forth distinct disclosures from the Fox Papers to meet the pool-
importance searching and pool-similarity searching elements. Pet. 22–23.
For instance, Fox SMART teaches using “general similarity computations,”
(Ex. 1008, 37–38) which Petitioner contends is pool-similarity searching, as
well as a “feedback” search loop in which results are ranked according to
importance to a user, and then further results are retrieved (id. at 24, Fig. 6),
which Petitioner contends is pool-importance searching. As we concluded
above, the feedback search function disclosed in the Fox Papers teaches
pool-importance searching, as required by claim 39.
For the foregoing reasons, Petitioner has shown by a preponderance of
the evidence that claims 28, 29, 30, 32, 34, and 39 of the ’352 patent are
unpatentable under 35 U.S.C. § 103(a) as they would have been obvious
over Fox Thesis, Fox SMART, and Fox Collection.
IPR2013-00478
Patent 5,544,352

32

D. Obviousness of Claims 26, 28–30, 32, 34, and 39 Over the Tapper
Papers
We instituted trial to determine whether claims 26, 28–30, 32, 34, and
39 are unpatentable under 35 U.S.C. § 103 as having been obvious over the
combined disclosures of Tapper 1976 and Tapper 1982 (collectively, “the
Tapper Papers”). Inst. Dec. 21–24.
We have considered Petitioner’s arguments and evidence, as well as
the counter-arguments in Patent Owner’s Response, and the evidence cited
therein, and conclude that Petitioner has not shown by a preponderance of
the evidence that each of claims 26, 28–30, 32, 34, and 39 of the ’352 patent
are unpatentable, under 35 U.S.C. § 103, as having been obvious over the
Tapper Papers.
1. Tapper 1976
Tapper 1976 discloses a “citation vector technique” for retrieving
legal information that seeks to overcome perceived deficiencies in Boolean
search strings. Ex. 1005, 270–71. Rather than characterizing a legal
document by the words it contains, vector matching focuses on the citations
the document contains. Id. at 263. Tapper 1976 also notes that the
technique may be used as an adjunct to a full-text retrieval system. Id. at
272.
By repeating the vector characterization of the documents, Tapper
1976 discloses that a matrix may be created that shows the similarities
between the documents. Id. By re-ordering the matrix, the documents may
be clustered according to their similarity. Id. The reference also discloses
that “second generation citations” may be used: “if a case cites cases A', B'
and C', and case A' cites a1', a2' and a3', case B' b1', b2' and b3' and case C'
IPR2013-00478
Patent 5,544,352

33

c1', c2' and c3' the original case would be represented by a combination of
its own vector, and those of cases A', B' and C'.” Id. at 266.
2. Tapper 1982
Tapper 1982 similarly focuses on the drawbacks of full-text searching
of legal documents and the alternative use of citation vectors for legal
research. Ex. 1006, 135–36. The reference discusses weighting certain
citation vectors more heavily than others, for example by the difference in
the ages of the citing and cited case. Id. at 138.
A pilot project implementing such a citation vector-based system is
also described by Tapper 1982. Id. at 139. The reference discloses a
correlation algorithm used in the pilot project to cluster together vectors with
a high degree of association. Id. at 143–44. Such clustering is said to permit
a document to be retrieved “not only because it is itself closely
associated with another target document, but also because both it and
the target document are closely associated with a third.” Id.
3. Claim 26
As discussed above, claim 26 requires steps of “creating a first
numerical representation for each identified object in the database based
upon the object’s direct relationship with other objects in the database.” We
find that this limitation is neither taught nor suggested by the combined
Tapper Papers.
Petitioner’s claim chart identifies several portions which allegedly
teach a first numerical representation. Pet. 47–48. For example, Tapper
1976 is cited as disclosing “quantifiable representation in the form of
numerical weighting” Id. (citing Ex. 1005, 263). These “quantifiable
representations” are of “other characteristics” of the citations, such as age or
IPR2013-00478
Patent 5,544,352

34

the importance of the court or jurisdiction deciding the case, not the citations
(relationships) themselves. Ex. 1005, 263.
Similarly, Tapper 1982 is cited as disclosing “[a]scription of
numerical values to vector elements.” Pet. 48 (citing Ex. 1006, 141). But
Tapper 1982 explicitly defines “vectors” as “the strings [a document]
contains and the frequency of their occurrence.” Ex. 1006, 134. In other
words, the “numerical values” of Tapper 1982 are the frequency of the
appearance of citation strings, which as we discussed above, connotes the
inclusion of letters. The Petition provides no citation to either Tapper Paper
that teaches representing direct relationships with a first numerical
representation.
In its Reply Brief, Petitioner identifies two other disclosures by the
Tapper Paper it contends satisfy the first numerical representation
limitation. First, Petitioner argues that “the legal citations in Tapper clearly
qualify as numerical representations.” Reply 4–5. The legal citations
Petitioner identifies, however, are in the exemplary form of “500 F.2d 411,”
which includes letters. As we have construed the term, this is not a
numerical representation, but rather the “strings” of the vectors discussed
above.
Second, Petitioner notes that the Tapper Papers describe assigning
cases in the database a unique ID number. Id. (citing Ex. 1006, 148). At
oral argument, Petitioner’s counsel directed our attention to Table 2 of
Tapper 1982, which includes in the leftmost column pairs of numbers which
signify pairs of documents. Tr. 14; Ex. 1006, 147. At most, the assignment
of these numbers could satisfy the marking step of claim 26 (Ex. 1006, 148
(“[t]he first column gives the numbers allocated to the cases.”)); they are not
IPR2013-00478
Patent 5,544,352

35

generating a first numerical representation. The document numbers
indicated by Petitioner are numerical representations of documents, not of
the relationships between those documents. Claim 26 requires that the first
numerical representations are based on direct relationships in the database.
The numbers allocated to the cases of Tapper 1982 cannot satisfy this
limitation.
Nor can the document number pairs of Table 2 be a first numerical
representation, as Tapper 1982 does not disclose that they represent a direct
relationship (i.e., one of the documents in the pair citing the second).
Rather, the pairs of documents appear to be listed together in the table
because of their high “correlation values.” Ex. 1006, 148. As Petitioner
acknowledges, these correlation values represent indirect relationships
between the documents (Reply 6 (“correlation values of cases’ indirect
relationships”)), therefore they cannot be a first numerical representation
that represents a direct relationship.
Petitioner argues in the alternative that “there is nothing non-obvious
about creating citation vectors consisting solely of numbers.” Reply 4–5.
At the outset, we note that this argument was presented for the first time in
the Reply; the sole modification to the Tapper Papers addressed in the
Petition is the combination of the disclosures of the two references. Pet. 45–
46. Nor did Petitioner present any testimony with the Petition regarding the
Tapper Papers, or how a person of ordinary skill in the art would have
modified the references. It would be a proper exercise of our discretion,
therefore, to not consider this argument and the Reply Declaration of Dr.
Fox (Ex. 1016), which presents testimony on the Tapper Papers for the first
IPR2013-00478
Patent 5,544,352

36

time.
3
See Office Patent Trial Practice Guide, 77 Fed. Reg. 48,756, 48,767
(Aug. 14, 2012) (“a reply that raises a new issue or belatedly presents
evidence will not be considered.”)
Even if we were to consider Petitioner’s Reply and Dr. Fox’s Reply
Declaration, however, we are not persuaded. Petitioner cites to various
portions of the Tapper Papers (Reply 4–5), but none of these citations
sufficiently establish a reason to substitute numerical representations for
those disclosed in Tapper. For example, Petitioner argues—using pieced-
together quotations—that “Tapper [1982] also makes clear that one could
‘very easily’ use a ‘simple conversion table’ to map ‘extracted’ citations to
any ‘chosen style.’” Id. at 5 (citing Ex. 1006, 136). Upon reading the full
context from which these quotes are drawn, however, it is clear that Tapper
1982 is discussing “parallel reports of the same decision.” Ex. 1006, 136.
In other words, Tapper 1982 does not contemplate converting letter-
containing case citations into numbers, but rather converting one letter-
containing citation into another.
Dr. Fox’s Reply Declaration (Ex. 1030 ¶¶ 107–115) relies on the same
arguments as Petitioner’s Reply, and we find them unpersuasive for the
same reasons. Nor are we persuaded by the portions of Dr. Jacobs’s cross-
examination Petitioner cites (Reply 5 (citing Ex. 1033, 313:7–316:23,
339:3–342:6)), as Dr. Jacobs’s testimony was to what a person of ordinary
skill would have understood from the ’352 patent specification, not the
Tapper Papers. See In re Vaeck, 947 F.2d 488, 493 (Fed. Cir. 1991)
(suggestion to make invention cannot “be founded . . . in the applicant’s

3
We address Patent Owner’s Motion to Exclude portions of the Reply
Declaration below.
IPR2013-00478
Patent 5,544,352

37

disclosure”). The record before us does not support the conclusion that a
person of ordinary skill in the art would have modified the combined
disclosures of the Tapper Papers to include a first numerical representation.
4. Dependent Claims
The remaining instituted claims all depend, directly or indirectly, from
claim 26, and thus incorporate claim 26’s requirement of a first numerical
representation. We, therefore, find that the Tapper Papers do not teach or
suggest all elements of these dependent claims.
E. Motion to Exclude
Patent Owner filed a Motion to Exclude (Paper 47) in which Patent
Owner seeks to exclude portions of the Reply Declaration of Dr. Edward A.
Fox (Ex. 1030) (“Reply Declaration”) submitted with Petitioner’s Reply. In
particular, Patent Owner identifies three issues with the Declaration, each of
which is based on the argument that portions of the Declaration are improper
reply evidence.
In its Reply, a Petitioner may only respond to arguments raised in the
Patent Owner’s Response. 37 C.F.R. § 42.23(a). “A reply that raises a new
issue or belatedly presents evidence will not be considered.” Office Patent
Trial Practice Guide, 77 Fed. Reg. at 48,767 (Aug. 14, 2012). The Practice
Guide provides, as indications of improper reply evidence, “new evidence
necessary to make out a prima facie case for . . . patentability or
unpatentability . . ., and new evidence that could have been presented in a
prior filing.” Id.
A motion to exclude evidence under 37 C.F.R. § 42.64(c), however,
“normally is not the proper vehicle for resolution of a dispute regarding
reply arguments and evidence exceeding the proper scope of a reply.” ABB,
IPR2013-00478
Patent 5,544,352

38

Inc. v. Roy-G-Biv Corp., Case IPR2013-00063, slip op. 13–14 (PTAB May
16, 2014) (Paper 71); Corning Inc. v. DSM IP Assets B.V., Case IPR2013-
00047, slip op 7 n.3 (PTAB May 1, 2014) (Paper 84) (characterizing such
motions as “now disfavored”). Rather, when evaluating the record after oral
argument, the Board is capable of determining what, if any, evidence
exceeds the proper scope of rely, and accordingly disregarding that
evidence.
While we, therefore, deny Patent Owner’s Motion, we also note that
even if it were proper, we would dismiss it as moot. With respect to the
objected-to portions of the Reply Declaration which discuss the Tapper
Papers, we have considered them above, found Dr. Fox’s testimony
unpersuasive, and found in favor of Patent Owner on the Tapper Papers
ground. With respect to the Fox Papers ground, we have found in favor of
Petitioner, but did not rely on any of the objected-to portions of the Reply
Declaration in so doing. A decision to exclude the Reply Declaration would,
therefore, not affect our determinations in this case.
F. Motions to Seal
Patent Owner filed a Motion to Seal (Paper 35) the Declaration of Dr.
Amy N. Langville (“Langville Declaration”) filed as Exhibit 2114.
Petitioner filed a Motion to Seal (Paper 42) the Transcript of the Deposition
of Amy N. Langville, Ph.D. (“Langville Transcript”) filed as Exhibit 1034.
Both of these motions are unopposed.
Regarding Patent Owner’s Motion to Seal, according to Patent Owner
paragraphs 25, 112, and 113 of the Langville Declaration makes reference to
certain facts about confidential licenses to the patents under review. Paper
IPR2013-00478
Patent 5,544,352

39

35, 3. Additionally, Patent Owner contends that this information has not
been made, and will not be made, public. Id.
Regarding Petitioner’s Motion to Seal, according to Petitioner, Patent
Owner has designated the transcript as confidential. Paper 42, 3. To avoid
public disclosure, therefore, Petitioner submits sealing the Langville
Transcript is appropriate. Id.
There is a strong public policy in favor of making information filed in
inter partes review proceedings open to the public. See Garmin Int’l v.
Cuozzo Speed Techs., LLC, Case IPR2012-00001 (PTAB Mar. 14, 2013)
(Paper 34). Under 35 U.S.C. § 316(a)(1), the default rule is that all papers
filed in an inter partes review are open and available for access by the
public.
4
The standard for granting a motion to seal is “good cause.”
37 C.F.R. § 42.54. A moving party bears the burden of showing that the
relief requested should be granted. 37 C.F.R. § 42.20(c).
Regarding Patent Owner’s Motion to Seal, Patent Owner, as the
moving party, has failed to carry its burden. Patent Owner identifies only
three paragraphs in the Langville Declaration that purportedly contain
confidential information. However, Patent Owner has not pointed to proof
in the record that any information contained in these paragraphs is
confidential. Additionally, although Patent Owner contends that this
information has not been made, and will not be made, public, Patent Owner
presented this information during the hearing on October 30, 2014, which

4
Additionally, we note that confidential information subject to a protective
order ordinarily would become public 45 days after final judgment in a trial.
Office Patent Trial Practice Guide, 77 Fed. Reg. at 48,761. However, after
denial of a petition to institute a trial or after final judgment in a trial, a party
may file a motion to expunge confidential information from the record. 37
C.F.R. § 42.56.
IPR2013-00478
Patent 5,544,352

40

was open to the public. See Tr. 54:12–25. We, therefore, determine that
Patent Owner has not met its burden of proof.
Regarding Petitioner’s Motion to Seal, Patent Owner’s designation of
the transcript as confidential is not sufficient to show that the transcript
contains confidential information. We, therefore, determine that Petitioner
has not met its burden of proof.
We recognize a denial of the motions to seal would unseal
immediately the material that Patent Owner desires to remain confidential
and the effect would be irreversible. Therefore, rather than denying the
motions at this time, we will provide Patent Owner and Petitioner one week
to (1) withdraw the motions to seal and request that we expunge Exhibits
2114 and 1034, or (2) withdraw the motions to seal, request that we expunge
Exhibits 2114 and 1034, and replace them with redacted versions that leave
out the confidential information. We note that we have not relied on the
three paragraphs of the Langville Declaration that Patent Owner identifies as
containing allegedly confidential information.
III. CONCLUSION
We conclude that Petitioner has shown by a preponderance of the
evidence that claims 26, 28–30, 32, 34, and 39 of the ’352 patent are
unpatentable under 35 U.S.C. § 103, as they would have been obvious over
Fox Thesis, Fox SMART, and Fox Collection, taken together.

IPR2013-00478
Patent 5,544,352

41

IV. ORDER
For the reasons given, it is
ORDERED that claims 26, 28–30, 32, 34, and 39 of U.S. Patent No.
5,544,352 are determined by a preponderance of the evidence to be
unpatentable;
FURTHER ORDERED that Patent Owner’s Motion to Exclude the
Reply Declaration of Dr. Edward A. Fox (Exhibit 1030) is denied;
FURTHER ORDERED that Exhibit 2114 and Exhibit 1034 will be
made available to the public after 5 PM Eastern five business days after the
entry date of this decision, unless prior to that time, each of Patent Owner
and Petitioner (1) withdraws the motions to seal and requests that we
expunge Exhibits 2114 and 1034, or (2) withdraws the motions to seal,
requests that we expunge Exhibits 2114 and 1034, and replaces them with
redacted versions that leave out the confidential information; and
FURTHER ORDERED that, because this is a final written decision,
parties to the proceeding seeking judicial review of the decision must
comply with the notice and service requirements of 37 C.F.R. § 90.2.
IPR2013-00478
Patent 5,544,352

42

FOR PETITIONER:

Heidi L. Keefe
COOLEY, LLP
hkeefe@cooley.com
dcpatentdocketing@cooley.com.

David Silbert
KEKER & VAN NEST, LLP
djs@kvn.com
efiling@kvn.com

FOR PATENT OWNER:

Martin M. Zoltick
Nancy J. Linck
Soumya P. Panda
ROTHWELL, FIGG, ERNST & MANBECK, P.C.
mzoltick@rfem.com
nlinck@rfem.com
SRA-IPR@rfem.com