Ex Parte Hagar et al

Patent Trial and Appeal Board

Apr 25, 2017

12263169 (P.T.A.B. Apr. 25, 2017)

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.
12/263,169 10/31/2008 David A. Hagar 078131.0106 7622
5073 7590 04/27/2017
RAKFR ROTTST T P
EXAMINER
2001 ROSS AVENUE LE, MICHAEL
SUITE 700
DALLAS, TX 75201-2980 ART UNIT PAPER NUMBER
2163
NOTIFICATION DATE DELIVERY MODE
04/27/2017 ELECTRONIC
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.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
ptomaill @bakerbotts.com
ptomai!2 @ bakerbotts .com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Ex parte DAVID A. HAGAR, PAUL A. JAKUBIK,
and STEPHEN S. JERNIGAN
Appeal 2016-007249
Application 12/263,169
Technology Center 2100
Before ALLEN R. MacDONALD, JOHN R. KENNY, and
MICHAEL J. ENGLE, Administrative Patent Judges.
MacDONALD, Administrative Patent Judge.
DECISION ON APPEAL
Appeal 2016-007249
Application 12/263,169
STATEMENT OF THE CASE
Appellants appeal under 35 U.S.C. § 134(a) from a rejection of claims
1—6, 8—27, and 29-43. Final Act. 1. We have jurisdiction under 35 U.S.C.
§ 6(b).
Exemplary Claim
Exemplary claims 1—3 under appeal read as follows (emphasis added
and multilevel numbering added (as used in the Final Action at page 3)):
1. A computerized method of determining latent relationships
in data comprising:
[a.] receiving a first matrix comprising a first plurality of terms,
the first matrix representing one or more data objects to be queried;
[b.] partitioning, before any Singular Value Decomposition
processing, the first matrix into a plurality of subset matrices by:
[i.] clustering similar vectors from the first matrix
together;
[ii.] creating a binary tree of clusters based on the
clustering of the similar vectors; and
[iii.] creating the plurality of subset matrices using the
created binary tree of clusters; and
[c.] processing each subset matrix with a natural language
analysis process to create a plurality of decomposed matrices
comprising:
[i.] a plurality of T0 matrices that provide a mapping of
the first plurality of terms into a first dimensional space;
[ii.] a plurality of So matrices that provide a scaling for
the plurality of To matrices; and
[iii.] a plurality of Do matrices that provide a mapping of
a plurality of documents into a second dimensional space;
[d.] determining a similarity between each of the plurality of To
matrices and a query from a user;
2
Appeal 2016-007249
Application 12/263,169
[e.] selecting, based on the determined similarities between
each of the plurality of To matrices and the query from the user,
a particular one of the plurality of To matrices that has the greatest
similarity to the query, and
[f.] generating a plurality of result terms using the selected To
matrix and the query.
2. The computerized method of determining latent
relationships in data of Claim 1, wherein partitioning the first matrix
into a plurality of subset matrices comprises:
forming each of the subset matrices so that each vector in the
first matrix appears in exactly one subset matrix, the size of each
subset matrix being a size that may be usefully processed by the
natural language analysis process
3. The computerized method of determining latent
relationships in data of Claim 1, wherein vectors are not discarded
from the first matrix prior to partitioning the first matrix into a
plurality of subset matrices.
Rejections
The Examiner rejected claims 1—6, 8, 9, 11—20, 22—27, 29, 30, 32—41,
and 43 under 35 U.S.C. § 103(a) as being unpatentable over the combination
of Behrens et al. (US 7,152,065 B2; iss. Dec. 19, 2006) and Sasaki et al.,
Web Document Clustering Using Threshold Selection Partitioning,
Proceedings ofNTCIR-4 (2004).1
1 Separate patentability is not argued for claims 4—6, 8, 9, 11—20, 22—27, 29,
30, 32-41, and 43. Except for our ultimate decision, these claims are not
discussed further herein.
3
Appeal 2016-007249
Application 12/263,169
The Examiner rejected claims 10, 21,31, and 42 under 35 U.S.C.
§ 103(a) as being unpatentable over the combination of Behrens, Sasaki, and
Roitblat et al. (US 2008/0059512 Al; pub. Mar. 6, 2008).2
Appellants ’ Contentions
l.A. Appellants contend that the Examiner erred in rejecting claim 1
under 35 U.S.C. § 103(a) because:
While the cited portions of Behrens may disclose selecting
sub-collections, they do not disclose “selecting... a particular
one of the plurality of To matricesand especially not selecting
one of the plurality of T0 matrices “that has the greatest
similarity to the query,” as recited in Claim 1.
App. Br. 11.
That is, while the cited portion of Behrens may disclose selecting
the best sub-collections based on rank, merely selecting sub­
collections does not disclose, teach, or suggest “selecting... a
particular one of the plurality of To matrices that has the
greatest similarity to the query,” as recited in Claim 1.
App. Br. 12.
Furthermore, the Examiner’s mapping of Claim 1 to
Behrens is flawed. For example, the Examiner maps Behrens'1
“sub-collections” to the claimed “subset matrices” (Final Office
Action, page 3, footnote 2, page 14, paragraph 47, and page 15,
paragraph 49) and then alleges that Behrens “discloses
decomposition of the sub-collections (i.e., subset matrices) into
the plurality of T0, So, and D0 matrices.” (Final Office Action,
page 14, paragraph 47.) The Examiner then collapses the
distinction between the claimed “subset matrices” and the
claimed “decomposed matrices comprising ... a plurality of T0
2 Separate patentability is not argued for claims 10, 21, 31, and 42. Thus,
the rejections of these claims turns on our decision as to claim 1. Except for
our ultimate decision, these claims are not discussed further herein.
4
Appeal 2016-007249
Application 12/263,169
matrices” by alleging that Behrens teaches that “a highest ranked
sub-collection (i.e., subset matrix that has a To with the greatest
similarity to the query) is selected.” (Final Office Action, page
15, paragraph 49.) However, selecting a sub-collection (equated
to the claimed “subset matrices” by the Examiner) does not
disclose selecting “a particular one of the plurality of T0
matrices,” as required by the above portion of Claim 1.
App. Br. 12 (emphasis omitted).
1 .B. Appellants contend that the Examiner erred in rejecting claim 1
under 35 U.S.C. § 103(a) because:
[T]he Examiner states in the Answer that the claimed “T0 matrix”
is created using SVD (Examiner’s Answer, Pages 3-4), but the
cited portions of Behrens for the disputed portion of Claim 1
(which require the claimed “T0 matrix”) are completely devoid
of any mention whatsoever of SVD.
Reply Br. 3 (footnote omitted).
Furthermore, other portions of Behrens specifically disclose that
the “term sets” of Behrens simply contain terms that correspond
to using k-means clustering, not SVD .... That is, Behrens
specifically defines the “term set” as simply being terms that are
the result of using k-means clustering, not SVD.
Reply Br. 6 (citing Behrens 5:62—6:5).
l.C. Appellants also contend that the Examiner erred in rejecting
claim 1 under 35 U.S.C. § 103(a) because:
[T]he Examiner provides no proof or explanation of how
Behrens' “term set” is a “T0 matrix” that was created by
processing a “subset matrix with a natural language analysis
process” and that “provide[sj a mapping of the first plurality of
terms into a first dimensional space," as required by other
portions of Claim 1.
Reply Br. 3.
5
Appeal 2016-007249
Application 12/263,169
2. Appellants also contend that the Examiner erred in rejecting claim 2
under 35 U.S.C. § 103(a) because:
The Examiner relies on 5:4-31 of Behrens as allegedly teaching
this portion of Claim 2. (Final Office Action, page 5.) However,
this is incorrect. While the cited portions of Behrens may
disclose partitioning a collection of data objects into sub­
collections, they do not disclose “forming each of the subset
matrices so that each vector in the first matrix appears in exactly
one subset matrix,” as recited in Claim 2.
App. Br. 13 (emphasis omitted).
In the Examiner’s Answer on Pages 5-7, the Examiner points to
new portions of both Behrens and Sasaki for this portion of
Claim 2 ... . Here, the Examiner jumps to the erroneous
conclusions that Behrens’ alleged disclosure of using clustering
to create homogenous sub-collections and Sasaki’s alleged
disclosure of disjoint clusters necessarily discloses “forming
each of the subset matrices so that each vector in the first matrix
appears in exactly one subset matrix,” as required by Claim 2.
However, this apparent reliance on inherency is incorrect.
Reply Br. 8 (Appellants’ emphasis omitted, Panel emphasis added).
3. Appellants also contend that the Examiner erred in rejecting claim 3
under 35 U.S.C. § 103(a) because:
[T]he proposed Behrens-Sasaki combination fails to disclose,
teach, or suggest “wherein vectors are not discarded from the
first matrix prior to partitioning the first matrix into a plurality of
subset matrices,” as recited in Claim 3. The Examiner relies on
4:59-67 and 5:1-3 of Behrens as allegedly teaching this portion
of Claim 3. (Final Office Action, page 5.) Specifically, the
Examiner states the following:
Preprocessing removes and ignores words (i.e.,
discards vectors) from the documents in the
collection (i.e., first matrix). The preprocessing
step is optional, which is interpreted to mean that
the invention can also work without preprocessing
(i.e., vectors are not discarded).
6
Appeal 2016-007249
Application 12/263,169
(Id. at footnote 4.) However, this is incorrect. While the cited
portions of Behrens may disclose a preprocessing step where
words may be ignored, they do not disclose “wherein vectors
are not discarded from the first matrix prior to partitioning the
first matrix into a plurality of subset matrices,” as recited in
Claim 3.
App. Br. 14—15 (bold-italicized emphasis added).
Furthermore, the Examiner’s footnote above suggesting that the
cited portion of Behrens discloses Claim 3 because it is an
optional step is flawed at least because other portions of Behrens
explicitly disclose that it is necessary to remove stop words for
the Behrens invention .... (Behrens, 6:25-35 . . .) That is,
Behrens explicitly states that “it is necessary to exclude high
frequency terms.” Id. Thus, the Examiner’s assertion that
removing terms from the Behrens invention is optional and
therefore “can also work without preprocessing” is incorrect.
App. Br. 15.
Issues on Appeal
Did the Examiner err in rejecting claims 1—3 as being obvious?
ANALYSIS
We have reviewed the Examiner’s rejections in light of Appellants’
Appeal Brief and Reply Brief arguments. We disagree with Appellants’
conclusions that the Examiner has erred. Rather, we concur with the
conclusions ultimately reached by the Examiner.
As to Appellants’ above contention l.A, Appellants appear to be
construing “selecting ... a particular one of the plurality of T0 matrices that
has the greatest similarity to the query” to require that the result of the
selection must be limited to exactly the matrix having the greatest similarity
to the query and cannot be a larger sub-collection containing a To matrix.
7
Appeal 2016-007249
Application 12/263,169
However, this argument does not adequately address the Examiner’s finding
that Behrens “determines the closest (i.e., greatest similarity) term set (i.e.,
T0 matrix) to the query vector” and “[t]he selected term set (i.e., To matrix)
is used to determine which sub-collection of documents to query.” Ans. 4.
Thus, even though each sub-collection may contain a To matrix, it is the To
matrix itself that is selected as having the greatest similarity. Ans. 4—5. We,
therefore, agree with the Examiner (Final Act. 4, 15) that Behrens suggests
this limitation, including at column 9, lines 2—33.
Further, we construe the limitation “selecting ... a particular one of
the plurality of T0 matrices that has the greatest similarity to the query” to
only require that the result of the selecting includes the matrix having the
greatest similarity. Claim 1 is an open-ended “comprising” claim that does
not preclude other unclaimed steps also selecting other less similar matrices
(e.g., the matrix with the second greatest similarity) so long as the greatest
similarity matrix is selected. David Netzer Consulting Eng V LLC v. Shell
Oil Co., 824 F.3d 989, 998 (Fed. Cir. 2016) (“[A] method claim with the
word ‘comprising’ appearing at the beginning generally allows for
additional, unclaimed steps in the accused process, but each claimed step
must nevertheless be performed as written.”).
As to Appellants’ above contention l.B, we disagree. Contrary to
Appellants argument that “Behrens specifically defines the ‘term set’ as
simply being terms that are the result of using k-means clustering, not SVD”
(Reply Br. 6, citing Behrens 5:62—6:5), Behrens states that k-means
clustering is performed (step 120) prior to step 130, which is the Singular
Value Decomposition (SVD), and again at step 140 on the reduced vector
8
Appeal 2016-007249
Application 12/263,169
spaces of the SVD to ultimately get the term set at step 160 (Behrens 5:4—
66). Thus, the term set is a result of the SVD.
As to Appellants’ above contention l.C, Appellants present in the
Reply Brief a new argument against the rejection of claim 1. Previously,
Appellants presented a different argument in the original Appeal Brief (App.
Br. 10—12), to which the Examiner responded (Ans. 3—5). In the Reply
Brief, however, Appellants further argue the “subset matrix with a natural
language analysis process” and “provide a mapping of the first plurality of
terms into a first dimensional space” limitations of claim 1. These
limitations were not previously argued in the Appeal Brief or raised by the
Examiner in the Answer. In the absence of a showing of good cause by
Appellants, we decline to consider an argument raised for the first time in
the Reply Brief, as the Examiner has not been provided a chance to respond.
See 37 C.F.R. § 41.41(b)(2) (2012); In re Hyatt, 211 F.3d 1367, 1373 (Fed.
Cir. 2000) (noting that an argument not first raised in the brief to the Board
is waived on appeal); Ex parte Nakashima, 93 USPQ2d 1834, 1837 (BPAI
2010) (informative) (explaining that arguments and evidence not timely
presented in the principal brief will not be considered when filed in a reply
brief, absent a showing of good cause explaining why the argument could
not have been presented in the principal brief); Ex parte Borden, 93 USPQ2d
1473, 1477 (BPAI 2010) (informative) (“Properly interpreted, the Rules do
not require the Board to take up a belated argument that has not been
addressed by the Examiner, absent a showing of good cause.”). Appellants
have provided no showing of good cause.
As to Appellants’ above contention 2, we disagree. Although we
agree with Appellants’ argument (App. Br. 13) that Behrens fails to disclose,
9
Appeal 2016-007249
Application 12/263,169
teach, or suggest each vector in the first matrix appears in exactly one subset
matrix, the Examiner correctly responds (Ans. 6) that Sasaki discloses
partitioning an original cluster into disjointed3 clusters. Appellants then
further argue “this apparent reliance on inherency is incorrect.” Reply Br. 8.
However, we see no relevance to this further argument as we find no
reliance on inherency by the Examiner.
As to Appellants’ above contention 3, we agree with Appellants’
argument (App. Br. 14) that Behrens at column 4, lines 59—67 and column 5,
lines 1—3 fails to disclose, teach, or suggest “wherein vectors are not
discarded from the first matrix prior to partitioning the first matrix into a
plurality of subset matrices.” However, we disagree with Appellants with
respect to Behrens at column 6, lines 25—35. Claim 3 only requires not
discarding prior to partitioning. The discarding step discussed at Behrens
column 6, lines 34—35 is an alternative to discarding during preprocessing
and is performed at the time the similarity is measured which is subsequent
to (i.e., not prior to) the partitioning step. The Examiner correctly points this
out. Ans. 8:3—11. Appellants do not further dispute this issue in the Reply
Brief.
3 In mathematics, the term “disjoint” means “(of two sets) having no
members in common; having an intersection that is empty.” E.J. Borowski
& Jonathan M. Borwein, HarperCollins Dictionary of Mathematics 169
(1991).
10
Appeal 2016-007249
Application 12/263,169
CONCLUSIONS
(1) The Examiner has not erred in rejecting claims 1—6, 8—27, and
29-43 as being unpatentable under 35 U.S.C. § 103(a).
(2) Claims 1—6, 8—27, and 29-43 are not patentable.
DECISION
The Examiner’s rejections of claims 1—6, 8—27, and 29-43 are
affirmed.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(l)(iv).
AFFIRMED
11