Ex Parte Clinchant et al

Patent Trial and Appeal Board

Jan 28, 2015

12233978 (P.T.A.B. Jan. 28, 2015)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
12/233,978 09/19/2008 Stephane Clinchant 20071423USNP-XER1890US01 3123
62095 7590 01/28/2015
FAY SHARPE / XEROX - ROCHESTER
1228 EUCLID AVENUE, 5TH FLOOR
THE HALLE BUILDING
CLEVELAND, OH 44115
EXAMINER
KINSAUL, DANIEL W
ART UNIT PAPER NUMBER
2165
MAIL DATE DELIVERY MODE
01/28/2015 PAPER
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
____________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________

Ex parte STEPHANE CLINCHANT and JEAN-MICHEL RENDERS
____________

Appeal 2012-011485
Application 12/233,978
Technology Center 2100
____________

Before BRUCE R. WINSOR, JOHN A. EVANS,
and JOSEPH P. LENTIVECH, Administrative Patent Judges.

LENTIVECH, Administrative Patent Judge.

DECISION ON APPEAL
Appellants1 appeal under 35 U.S.C. § 134(a) from the final rejection
of claims 2–11 and 13–20. We have jurisdiction under 35 U.S.C. § 6(b).
Claims 1 and 12 are cancelled.
We reverse.

1 The real party in interest identified by Appellants is Xerox Corporation.
App. Br. 1.
Appeal 2012-011485
Application 12/233,978

2
STATEMENT OF THE CASE
Appellants’ invention is directed to information retrieval systems.
Spec. ¶ 1. Claim 8, which is illustrative, reads as follows:
8. A multimedia information retrieval method performed by
an electronic device, the method comprising:
performing an initial information retrieval process
respective to a multimedia reference repository to return a set of
initial repository documents;
computing values of at least one monomodal pairwise
similarity measure for candidate documents of the multimedia
reference repository respective to repository documents of the
set of initial repository documents; and
identifying a set of top-ranked documents of the
multimedia reference repository based at least in part on the
values computed for the candidate documents;
wherein the computing is performed before the
performing of the initial information retrieval process, and
the method further comprises storing the computed
values in a storage of or accessible by the electronic device,
the identifying comprising selecting and retrieving the
values computed for the candidate documents from the storage.
App. Br. 1718.
References
The Examiner relies on the following prior art in rejecting the claims:
Chen et al. US 6,728,752 B1 April 27, 2004
(hereinafter “Chen”)
Peh US 2006/0041604 A1 February 23, 2006

Rejections
Claims 2, 3, 5–11, and 13–20 stand rejected under 35 U.S.C. § 102(b)
as being anticipated by Chen. App. Br. 6.
Appeal 2012-011485
Application 12/233,978

3
Claim 4 stands rejected under 35 U.S.C. § 103(a) as being
unpatentable over Chen and Peh. Id. at 14.

Rather than repeat the arguments here, we refer to the Appeal Brief
(“App. Br.” filed March 20, 2012) and the Reply Brief (“Reply Br.” filed
July 31, 2012) for the positions of Appellants and the Examiner’s Answer
(“Ans.” mailed May 31, 2012) for the positions of the Examiner. Only those
arguments actually made by Appellants have been considered in this
decision. Arguments that Appellants did not make in the Briefs have not
been considered and are deemed to be waived. See 37 C.F.R.
§ 41.37(c)(1)(iv) (2011).

DISPOSITIVE ISSUE
Does Chen disclose “the identifying comprising selecting and
retrieving the values computed for the candidate documents from the
storage” (hereinafter the “selecting and retrieving limitations”), as recited in
claim 8?

ANALYSIS
CLAIM 8
The Examiner finds:
The computing is performed before the performing of the initial
information retrieval process; storing the computed values in a
storage of or accessible by the electronic device, the identifying
comprising selecting and retrieving the values computed for the
candidate documents from the storage is disclosed at col. 10,
lines 9-18 (i.e. feature vectors for each document are stored in a
database where they are correlated with the documents they
Appeal 2012-011485
Application 12/233,978

4
correspond to); col. 13, lines 10-26 (i.e. once text vectors have
been calculated, the similarity between two text vectors can be
calculated via a simple cosine distance; the cosine distances
between pairs of documents can be used to cluster documents
based on features); . . . col. 20, lines 66-67 (i.e. clusters of
documents created as described above are used in a system for
searching); col. 22, lines 33-47 (i.e. preprocessing step used to
pre-compute information needed during browsing and provide
the initial organization of data; a set of distinct features is pre-
computed for each document stored as vectors . . . the
documents are clustered into groups based on each of these
features); col. 22, lines 48-57 (i.e. after a user begins a search
query, the pre-computed text clusters are ranked in terms of
relevance to the query terms using the cosine distance).
Ans. 5–6 (emphasis omitted). The Examiner further finds Chen’s clusters
are the result of monomodal pairwise similarity comparisons and “represent
a ‘similarity measure’ for purposes of meeting the limitations” and “[t]he
limitation language of ‘computing values’ of a ‘similarity measure’ is broad
enough to encompass the resulting clusters that are the product of similarity
comparisons.” Id. at 14.
Appellants acknowledge Chen discloses computing monomodal
pairwise similarity measure values when forming clusters. App. Br. 8.
Appellants contend, however, Chen does not disclose storing the computed
pairwise similarity measure values, but instead, discloses storing, for each
document, an annotation indicating a cluster to which the document belongs.
Id. at 9. According to Appellants, Chen explicitly states when pieces of
information are stored for later use (e.g., feature vectors, histogram
representations, and cluster results); Chen does not explicitly state that the
pairwise similarity measure values are stored; and, therefore, Chen cannot
disclose the selecting and retrieving limitations. Reply Br. 5 (citing Chen,
Appeal 2012-011485
Application 12/233,978

5
Fig. 3, col., 10, ll. 14–15; col. 11, ll. 15–17; col. 16, ll. 50–51; col. 22, ll. 37–
39; and col. 23, ll. 27–28.). Appellants further contend:
Moreover, the cluster results are not equivalent or
analogous to the pairwise similarity measures of the present
claims. The cluster centroid is not a similarity measure at all,
but rather identifies the mean location of a set (i.e. cluster) of
document feature vectors. Distances between documents and
cluster centroids are not stated as being precomputed [sic],
stored, and used in the IR [information retrieval], but even if
such distances were stored and used they would not be pairwise
similarity measure values but rather would be distances
between entities of two different kinds, i.e. a document and a
centroid.
Id. (emphasis omitted).
We find Appellants’ contentions persuasive. Chen discloses “‘k-
means’ clustering is used to assign objects [i.e., documents] to k different
clusters.” Chen, col. 19, ll. 39–40. Chen further discloses k-means
clustering begins with selecting k objects as cluster centers (id. at col. 19, ll.
43–45); computing each object’s similarity with respect to each of the k
objects selected as cluster centers (id. at col. 20, ll. 44–45); and assigning
each object to a cluster having a center with which the object is most similar
(id. at col. 19, ll. 43–45). Chen also discloses that the clusters are computed
prior to returning an initial set of documents in response to a search query.
Id. at col. 22, ll. 54–57. Chen, therefore, discloses “wherein the computing
is performed before the performing of the initial information retrieval
process,” as recited in claim 8.
Further, Chen discloses, inherent in the computation of the initial
clusters, storing, at least temporarily, at least one computed monomodal
pairwise similarity measure value, computed for one cluster center, for
subsequent comparison to another monomodal pairwise similarity measure
Appeal 2012-011485
Application 12/233,978

6
value, computed for another cluster center, in order to determine the cluster
center to which the object is most similar. As such, Chen discloses “storing
the computed values in a storage of or accessible by the electronic device,”
as recited in claim 8.
Chen discloses after the initial assignment of objects to the closest
cluster center, the “cluster centers are recomputed as the means of their
members [i.e., the vector centroid of the objects in a cluster (see Chen, col.
20, ll. 16–20)]” and “[t]he process of (re)assignment of objects and re-
computation of means is repeated several times until it converges.” Chen,
col. 19, ll. 43–45. Chen, therefore, discloses the stored clusters are the result
of monomodal pairwise similarity comparisons of each document to a vector
centroid. Claim 8 requires the values of the at least one monomodal
pairwise similarity measure for candidate documents of the multimedia
reference repository to be computed respective to repository documents of
the set of initial repository documents. See App. Br., Claims Appendix. As
such, Chen’s selection and retrieval of clusters relevant to a search query
(see Chen, col. 22, ll. 54–57) does not teach or suggest the selecting and
retrieving limitations.
We do not reach Appellants’ further allegations of error because we
find the issue discussed above to be dispositive.
Accordingly, we do not sustain the Examiner’s rejections of (1) claim
8; (2) independent claims 16 and 18, which recite similar limitations; and (3)
claims 2–7, 911, 13–15, 17, 19, and 20, which depend variously from
claims 8, 16, and 18.

Appeal 2012-011485
Application 12/233,978

7
DECISION
The decision of the Examiner to reject claims 2–11 and 13–20 is
reversed.

REVERSED

llw