Ex Parte Sarwar et al

Patent Trial and Appeal Board

Mar 20, 2013

11646900 (P.T.A.B. Mar. 20, 2013)

UNITED STATES PATENT AND TRADEMARK OFFICE
____________________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
____________________

Ex parte BADRUL M. SARWAR and JOHN A. MOUNT
____________________

Appeal 2010-006668
Application 11/646,900
Technology Center 2100
____________________

Before CARL W. WHITEHEAD, JR., ERIC S. FRAHM, and
ANDREW J. DILLON, Administrative Patent Judges.

FRAHM, Administrative Patent Judge.

DECISION ON APPEAL

Appeal 2010-006668
Application 11/646,900

2
STATEMENT OF CASE1
Introduction
Appellants appeal under 35 U.S.C. § 134(a) from a final rejection of
claims 1-30. We have jurisdiction under 35 U.S.C. § 6(b).
We affirm-in-part.
Appellants’ Invention
Appellants’ invention relates to the field of data analysis. Appellants
disclose methods and systems for automatic segmentation of unstructured
description based on a header token (i.e., a unit of text that has meaning
according to lexical rules) (Spec. ¶¶ [0016] and [0017]; Fig. 4). A
probability of token relevance and irrelevance are determined for each token
in the description based on header tokens (Spec. ¶ [0018]; claim 1).
Different segments of the description (i.e., different sequences of tokens) are
selected and probabilities of relevance and irrelevance are determined for the
selected segment. The most relevant segment is used as an index to the
descriptions (Spec. ¶ [0020]; Fig. 5; claim 20).
Exemplary Claims
Exemplary independent claims 1 and 20 under appeal, with emphasis
added to the disputed portions of the claims, read as follows:
1. A system comprising:

a first machine to transmit a header and an unstructured description
corresponding to the header; and

1 Throughout our decision, we make reference to the Appeal Brief filed
December 7, 2009 (“App. Br.”); the Examiner’s Answer mailed January 27,
2010 (“Ans.”); and the Reply Brief filed March 29, 2010 (“Reply Br.”).
Appeal 2010-006668
Application 11/646,900

3
a second machine in at least selective communication with the first
machine, the second machine to,

receive the header and the unstructured description,

determine relevant state values and irrelevant state values
corresponding to each of a plurality of tokens in the unstructured
description, the determining based on the header,

indicate as most relevant a set of tokens based, at least in part,
on the relevant state values of the set of tokens and the irrelevant state values
of those of the plurality of tokens outside of the set of tokens, and

store the set of tokens indicated as most relevant on a machine-
readable storage medium.

20. A method of automatic text segmentation, the method comprising:
estimating, for each token in a set of tokens in a description, through
use of a machine connected to a network, a probability that the token is
irrelevant;

associating, with each token in the set of tokens in the description, one
of a first value, a second value, or a third value, based on whether,
respectively, the token occurs in a header of the description, a lexical
association exists between the token and a token in the header, or the lexical
association is absent and the token does not occur in the header;

iterating, through use of the machine, over a plurality of groups of
sequential tokens in the description, in each iteration, selecting a group,
computing a relevance value of the selected group based, at least in
part, on at least one estimated probability of one or more tokens outside the
selected group and on values associated with one or more tokens in the
selected group; and

indicating, through use of the machine, one of the plurality of groups
as having a greatest relevance value.
Appeal 2010-006668
Application 11/646,900

4
Examiner’s Rejection
The Examiner rejected claims 1-30 as being unpatentable under
35 U.S.C. § 103(a) over Mark Styvers et al., Probabilistic Author-Topic
Models for Information Discovery, 10TH ACM SigKDD Conference
Knowledge Discovery and Data Mining (Seattle, 2004), pp. 306-15 (August
22-25, 2004) (hereinafter, “Steyvers”) in view of Chen (U.S. Pat. Pub.
2005/0165753 A1). Ans. 4-20.
Appellants’ Contentions
(1) With regard to claim 12, rejected under 35 U.S.C. § 103(a) as
being unpatentable over Steyvers and Chen, Appellants make numerous
arguments (App. Br. 16-19; Reply Br. 2-3), including:
(a) The portions of Chen cited by the Examiner do not teach
determining relevant and irrelevant state values based on a header
(App. Br. 16-18);
(b) The “subweb” of Chen is not functionally equivalent to
“header” of claim 1 and the Appellants’ specification (App. Br. 17-
18);

2 Separate patentability is not argued for claims 2-4 depending from claim 1
(App. Br. 19). We select claim 1 as representative of the group of claims 1-
4; pursuant to our authority under 37 C.F.R. § 41.37(c)(1)(vii). See In re
McDaniel, 293 F.3d 1379, 1383 (Fed. Cir. 2002) (“If the brief fails to meet
either requirement [of 37 CFR § 1.192(c)(7)], the Board is free to select a
single claim from each group of claims subject to a common ground of
rejection as representative of all claims in that group and to decide the
appeal of that rejection based solely on the selected representative claim.”)
In view of the foregoing, our analysis regarding this group will only address
the merits of representative claim 1.

Appeal 2010-006668
Application 11/646,900

5
(c) There is no discussion of “state values” in Steyvers and
Chen (Reply Br. 2-3); and
(d) The topic-word and author-topic distributions of Steyvers
are not functionally equivalent to “a header and an unstructured
description corresponding to the header” as recited in claim 1 (App.
Br. 18-19; Reply Br. 3).
(2) With regard to claims 5 and 193 rejected under 35 U.S.C. § 103(a)
as being unpatentable over Steyvers and Chen, Appellants argue that neither
Chen nor Steyvers teach estimating a first probability that a token occurs as
an irrelevant token in a description and a second probability that the token
occurs as a relevant token in the description, based, at least in part, on a
header of the description as required by claims 5 and 19 (App. Br. 19-21;
Reply Br. 3-4).
(3) With regard to the rejection of claims 20 and 244 under 35 U.S.C.
§ 103(a) as being unpatentable over Steyvers and Chen, Appellants argue
(App. Br. 21-23; Reply Br. 4-5) that:

3 Separate patentability is not argued for claims 6-18 depending from claim
5 (App. Br. 21). Appellants group the subject matter common to claims 5
and 19 in their arguments (App. Br. 19). We select claim 5 as representative
of the group of claims 6-19 pursuant to our authority under 37 C.F.R. §
41.37(c)(1)(vii). In view of the foregoing, our analysis regarding this group
will only address the merits of representative claim 5.
4 Separate patentability is not argued for claims (i) 21-23 depending from
claim 20 (App. Br. 23) and (ii) 25 and 26 depending from claim 24 (Id.).
Appellants group the subject matter common to claims 20 and 24 in their
arguments (App. Br. 21). Claims 20 and 24 contain similar limitations. In
view of the foregoing, our analysis will only address the merits of
independent claims 20 and claim 24, from which claims 21-23, 25, and 26
respectively depend.
Appeal 2010-006668
Application 11/646,900

6
(a) The portions of Chen cited by the Examiner do not teach
associating a value based on whether a token occurs in a header, a
lexical association exists, or the lexical association is absent; as taught
by claims 20 and 24 (id.);
(b) A site of Chen is not a token (Reply Br. 5); and
(c) Steyvers does not teach associating a value with a token
(Reply Br. 5).
(4) With regard to claims 27 and 295 rejected under 35 U.S.C. §
103(a) as being unpatentable over Steyvers and Chen, Appellants argue that
the portions of Chen cited by the Examiner do not mention estimated
probabilities of irrelevance and estimated probabilities of relevance (App.
Br. 23-24; Reply Br. 5), and neither Steyvers nor Chen disclose estimated
probabilities of irrelevance and estimated probabilities of relevance (App.
Br. 24).
Issues on Appeal
Based on Appellants’ arguments in the Appeal Brief (App. Br. 16-25)
and the Reply Brief (Reply Br. 2-5), the following issues are presented on
appeal:
(1) Did the Examiner err in rejecting claims 1-4 because the
combination of Steyvers and Chen fails to disclose the limitations at issue in
representative claim 1?

5 Separate patentability is not argued for claims (i) 28 depending from claim
27 (App. Br. 24) and (ii) 30 depending from claim 29 (Id.). Appellants
group the subject matter common to claims 27 and 29 in their arguments
(App. Br. 23). We select claim 27 as representative of the group of claims
27-30 pursuant to our authority under 37 C.F.R. § 41.37(c)(1)(vii). In view
of the foregoing, our analysis regarding this group will only address the
merits of representative claim 27.
Appeal 2010-006668
Application 11/646,900

7
(2) Did the Examiner err in rejecting claims 5-19 because the
combination of Steyvers and Chen fails to teach (a) estimating a first
probability that a token occurs as an irrelevant token in a description, and (b)
a second probability that the token occurs as a relevant token in the
description, based, at least in part, on a header of the description, as required
by representative claim 5?
(3) Did the Examiner err in rejecting claims 20-26 because the
combination of Steyvers and Chen fails to teach or suggest associating a
value based on whether a token occurs in a header, a lexical association
exists, or the lexical association is absent, as required by independent claims
20 and 24, and claims 21-23, 25, and 26 depending therefrom?
(4) Did the Examiner err in rejecting claims 27-30 because Chen, and
thus the combination of Steyvers and Chen, fails to teach the estimated
probabilities of irrelevance and estimated probabilities of relevance, as
required by representative claim 27?

ANALYSIS
We have reviewed the Examiner’s rejections in light of Appellants’
contentions in the Appeal Brief (App. Br. 16-25) and the Reply Brief (Reply
Br. 2-5) that the Examiner has erred.
We agree with Appellants’ contentions (App. Br. 21-23; Reply Br. 4-
5) with respect to the obviousness rejection of independent claims 20 and 24
over Steyvers and Chen as applied by the Examiner (see Ans.13-15 and 23-
24). However, with regard to the obviousness rejection of the representative
claims 1, 5, and 27 over the combination of Steyvers and Chen, we agree
with the Examiner (Ans. 5-10, 16-17, and 20-24) that the combination of
Appeal 2010-006668
Application 11/646,900

8
Steyvers and Chen teaches or suggests the respective limitations of
representative claims 1, 5, and 27.
Issue 1 - Rejection of Representative Claim 1:
With regard to representative claim 1, we adopt as our own (i) the
findings and reasons set forth by the Examiner in the action from which this
appeal is taken (Ans. 5-7), and (ii) the reasons set forth by the Examiner in
the Examiner’s Answer in response to the Appellants’ Appeal Brief (Ans.
20-22). We highlight and amplify certain teachings and suggestions of the
references as follows.
We agree with the Examiner’s findings (Ans. 5 and 20-22) that
Steyvers teaches (i) receiving topic words and abstracts (Steyvers, Abs.; §§
3.1 and 3.2), and (ii) assigning relevance to a plurality of word tokens in the
abstracts based on topic words (Steyvers, § 3.1). We also agree with the
Examiner’s findings (Ans. 6 and 21) that Chen teaches assignment of
relevance and irrelevance of a site to a particular topic (Chen, ¶¶ [0036] and
[0037]). Further, we agree with the Examiner’s determination (Ans. 6) that
it would have been obvious to combine Steyvers and Chen based on the
rationale set forth by the Examiner, including using the relevance and
irrelevance values of Chen in the relevance assignment of Steyvers to a
plurality of word tokens in the abstracts based on topic words, to facilitate
topical categorization of domains and/or paths thereby providing focused
results.
We note that each reference cited by the Examiner must be read, not
in isolation, but for what it fairly teaches in combination with the prior art as
a whole. See In re Merck & Co., 800 F.2d 1091, 1097 (Fed. Cir. 1986) (one
Appeal 2010-006668
Application 11/646,900

9
cannot show non-obviousness by attacking references individually where the
rejections are based on combinations of references) (citation omitted).
Appellants make arguments (App. Br. 16-18; Reply Br. 2-3) with
regard to representative claim 1 concerning the individual shortcomings in
the teachings of each of the applied references to Steyvers and Chen. We
have carefully reviewed these arguments, however, they are not convincing
of the non-obviousness of the claimed limitation set forth in Appellants’
Contention (1) (a). For example, Appellants argue (App. Br. 16-18; Reply
Br. 2-3) that the cited portions of Chen and Steyvers, taken individually, do
not teach determining relevant and irrelevant state values based on a header
as recited in representative claim 1, when the Examiner has relied (Ans. 5-7)
upon the combination of Steyvers and Chen to teach this limitation.
Appellants argue (App. Br. 17-18) that the “subweb” of Chen is not
functionally equivalent to the “header” of representative claim 1 and the
Appellants’ specification (Spec. ¶ [0017]), when the Examiner has relied
(Ans. 5) upon Steyvers for disclosing the header (Steyvers, Abs.; §§ 3.1 and
3.2).
Furthermore, Appellants assert (Reply Br. 2-3) that there is no
discussion of “state values” in Steyvers and Chen. Chen teaches assigning
weights to sites to illustrate relevance of a site to a topic (Chen, ¶¶ [0036]
and [0037]) which is considered by the Examiner (Ans. 7 and 21) to teach
relevance and irrelevance state values.
Appellants contend (Reply Br. 3) that topic-word and author-topic of
Steyvers (i) refer to probability distributions, and (ii) are not functional
equivalents to “a header and an unstructured description corresponding to
the header” as recited in representative claim 1. We agree with the
Appeal 2010-006668
Application 11/646,900

10
Examiner (Ans. 5) that Steyvers teaches a header and unstructured
description corresponding to the header since the distributions of Steyvers
are obtained from the topics and the abstracts from the CiteSeer database
(Steyvers, §§ 3.1 and 3.2). Steyvers teaches the association of authors to
topics in §3.2. Steyvers considers the authors who tend to produce the most
words for that topic in the CiteSeer abstracts and computes distributions of
probabilities for authors and the words (Steyvers, Fig. 2).
In view of the foregoing, we will sustain the Examiner’s obviousness
rejection of representative claim 1, as well as claims 2-4 grouped therewith,
over the combination of Steyvers and Chen.
Issue 2 - Rejection of Representative Claim 5:
With regard to representative claim 5, we adopt as our own (i) the
findings and reasons set forth by the Examiner in the action from which this
appeal is taken (Ans. 8-10), and (ii) the reasons set forth by the Examiner in
the Examiner’s Answer in response to the Appellants’ Appeal Brief (Ans.
22-23). We highlight and amplify certain teachings and suggestions of the
references as follows.
We agree with the Examiner’s findings (Ans. 9 and 22-23) regarding
representative claim 5 that Steyvers teaches probability that a token exists
based on a topic word (see Steyvers, Abstract, pp. 306-08). We also agree
with the Examiner that Chen teaches the assignment of relevance and
irrelevance weights to sites based upon topics (Chen, ¶¶ [0036] and [0037]).
We concur with the Examiner’s reliance upon the combination of Steyvers
and Chen as teaching or suggesting estimating probabilities of relevance and
irrelevance based on a header, for the reasons set forth by the Examiner (see
Ans. 10).
Appeal 2010-006668
Application 11/646,900

11
Appellants’ second contention (see Appellants’ Contention (2) supra)
regarding representative claim 5 is not persuasive because Steyvers teaches
assigning probabilities that a token exists for a topic, and Chen teaches
relevance and irrelevance weights based upon topics. Therefore, we agree
with the Examiner (Ans. 10) that it would have been obvious to combine the
teachings of Steyvers and Chen to perform a method of automatic text
segmentation using probabilities of relevance and irrelevance to present
better focused results for domain-specific searches, as set forth in
representative claim 5.
In view of the foregoing, we will sustain the Examiner’s obviousness
rejection of representative claim 5, as well as claims 6-19 grouped therewith,
over the combination of Steyvers and Chen.
Issue 3 - Rejection of Independent Claims 20 and 24:
We agree with Appellants’ third contention (see Appellants’
Contentions (3)(a)-(c) supra) as to independent claims 20 and 24. Claims 20
and 24 require associating with each token, a value based on whether (a) a
token occurs in a header, (b) a lexical association exists, or (c) a lexical
association is absent (see claims 20 and 24).
With regard to Appellants’ Contention (3)(a) supra, we do not agree
with the Examiner (Ans. 23) that Chen’s assignment of weights to sites
based on relevance of a site to the topic of the subweb corresponds to
associating with each token, a value based on (a), (b) or (c) as indicated
above with respect to independent claims 20 and 24.
The Examiner, either in the rejection (Ans. 13-14) or in the response
(Ans. 23-24) to Appellants’ contention, failed to describe how the relevance
weights of Chen correspond to associating a value based on whether (a) a
Appeal 2010-006668
Application 11/646,900

12
token occurs in a header, (b) a lexical association exists, or (c) a lexical
association is absent. Further, the Examiner relied upon Steyvers to teach
relevance to a header (Ans. 23), and concluded that using relevant and
irrelevant state values of Chen (Chen, ¶¶ [0037], [0048], and [0073]) could
be used in the system of Steyvers for the purpose of facilitating topical
categorization, thereby providing focused results (Ans. 17).
With regard to Appellants’ Contention (3)(b) supra, we agree with
Appellants’ contention that a site of Chen is not a token. The Examiner has
established (Ans. 13-14 and 23-24) that the word tokens of Steyvers
(Steyvers, Abstract; pp. 306-08) can be interpreted as tokens as required by
independent claims 20 and 24. In addition, Chen was used by the Examiner
to explain relevance and irrelevance weights (see Ans. 14 and 23).
With regard to Appellants’ Contention (3)(c) supra, we do not agree
with Appellants that Steyvers does not teach associating a value with a token
because the Examiner used the probabilistic distribution over words for a
topic of Steyvers (see Steyvers, Abstract; pp. 306-08) to teach the
association of value with a token as required by independent claims 20 and
24 (see Ans. 23-24).
Accordingly, since the Examiner has not established a prima facie
case of obviousness for independent claims 20 and 24, or claims 21-23, 25,
and 26 depending therefrom, with respect to the combination of Steyvers
and Chen, we will not sustain the Examiner’s obviousness rejection of
claims 20-26.
Issue 4 - Rejection of Representative Claim 27:
With regard to representative claim 27, we adopt as our own (i) the
findings and reasons set forth by the Examiner in the action from which this
Appeal 2010-006668
Application 11/646,900

13
appeal is taken (Ans. 16-17), and (ii) the reasons set forth by the Examiner
in the Examiner’s Answer in response to the Appellants’ Appeal Brief (Ans.
24). We highlight and amplify certain teachings and suggestions of the
references as follows.
We agree with the Examiner’s findings (Ans. 16-17 and 24) regarding
representative claim 27 that Chen teaches filtering the sites that are known to
be irrelevant a priori to the topic and assigning relevance weights to sites
that are not filtered (see Chen, ¶ [0073]). Because (i) Chen’s subweb
generator 1008 uses a stop list 1020 which includes a list of sites that are not
relevant to a topic or characteristic of interest (i.e., irrelevant) to
automatically “disclude[]” from the list of sites from being generated in
subweb (Chen,¶ [0073]), and (ii) certainty is the highest form of probability
(i.e., the estimated probability of relevance is 100%, and of irrelevance is
0%), Chen’s a priori irrelevance determination is encompassed by the
recitation in representative claim 27 of “an estimated probability of
irrelevance and an estimated probability of relevance.” Appellants’
arguments (App. Br. 24; Reply Br. 5) to the contrary are not persuasive.
Appellants also argue (App. Br. 23-24; Reply Br. 5) that the portions
of Chen cited by the Examiner do not mention estimated probabilities of
irrelevance and estimated probabilities of relevance as recited by
representative claim 27 (Chen, Fig.5; ¶¶ [0036], [0037], [0048], [0073], and
[0075]). After careful review of these arguments, we find that the estimation
of probabilities of irrelevance and relevance are taught by Chen since Chen
teaches (i) “disclud[ing]” from the subweb, sites that are known a priori to
be irrelevant to a topic (i.e., estimation of probabilities of irrelevance), and
(ii) assigning weights to sites that are not filtered out (i.e., estimation of
Appeal 2010-006668
Application 11/646,900

14
probabilities of relevance) (Chen, ¶ [0073]). Appellants’ arguments (App.
Br. 24; Reply Br. 5) that Chen does not teach estimated probabilities are also
not persuasive because Chen (i) was relied upon to teach relevance
weights/values (Ans. 17 and 24), and (ii) teaches or suggests estimated
probabilities by limiting sites to only sites that are relevant as discussed
supra.
We agree with the Examiner (Ans. 16-17) that the combination of
Steyvers (Steyvers, pp. 308 and 309) and Chen (Chen, ¶ [0073]) teaches or
suggests estimated probabilities of relevance and irrelevance based on a set
of one or more tokens in a header of the description as required by
representative claim 27, for the reasons set forth by the Examiner (Ans. 17).
Appellants’ argument (App. Br. 24) that neither Styvers nor Chen discloses
estimated probabilities of irrelevance and relevance is not persuasive in light
of the facts that (i) Styvers (pp. 308 and 309; Fig. 2)6 teaches estimated
probabilities, and (ii) Appellants admit with regard to the arguments
presented regarding the rejection of claim 1 that Steyvers teaches estimating
probabilities (see App. Br. 18-19, stating that Styvers relates to “probability
distributions”).
In view of the foregoing, we will sustain the Examiner’s obviousness
rejection of representative claim 27, as well as claims 28-30 grouped
therewith, over the combination of Steyvers and Chen.

6 Notably, Appellants attention has been directed by the Examiner to Figure
2 and page 309 of Styvers regarding regarding claim 1 (Ans. 5), to pages
308-09 of Styvers regarding claim 5 (Ans. 8-9), and to pages 308-09 of
Styvers regarding claim 27 (Ans. 16).
Appeal 2010-006668
Application 11/646,900

15

CONCLUSIONS
(1) The Examiner has not erred in in rejecting representative claim 1
on appeal, and claims 2-4 grouped therewith, since Steyvers and Chen teach
the limitations of representative claim 1.
(2) Because the combination of Steyvers and Chen teach estimating a
first probability that a token occurs as an irrelevant token in a description
and a second probability that the token occurs as a relevant token in the
description, based, at least in part, on a header of the description as required
by representative claim 5, the Examiner has not erred in rejecting claim 5
and claims 6-19 grouped therewith.
(3) Appellants have adequately shown that the Examiner has erred in
rejecting independent claims 20 and 24, as well as claims 21-23, 25, and 26
depending therefrom, because the combination of Steyvers and Chen fails to
teach or suggest associating a value based on whether a token occurs in a
header, a lexical association exists, or the lexical association is absent, as
recited in independent claims 20 and 24.
(4) The Examiner has not erred in rejecting claims 27-30 because the
combination of Steyvers and Chen teaches or suggests the estimated
probabilities of irrelevance and estimated probabilities of relevance required
by representative claim 27, and claims 28-30 grouped therewith.

DECISION
The Examiner’s rejection of claims 1-19 and 27-30 under
35 U.S.C. § 103(a) is affirmed.
Appeal 2010-006668
Application 11/646,900

16
The Examiner’s rejection of claims 20-26 under 35 U.S.C. § 103(a) is
reversed.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED-IN-PART

tkl