Ex Parte Kirpal

Patent Trial and Appeal Board

Jun 28, 2016

12013289 (P.T.A.B. Jun. 28, 2016)

UNITED STA TES p A TENT AND TRADEMARK OFFICE
APPLICATION NO. FILING DATE
12/013,289 0111112008
76239 7590 06/30/2016
Weaver Austin Villeneuve & Sampson - YAHI
Attn: Yahoo!
P.O. BOX 70250
OAKLAND, CA 94612-0250
FIRST NAMED INVENTOR
Alok S. Kirpal
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
ATTORNEY DOCKET NO. CONFIRMATION NO.
YAH1Pl31/Y04033USOO 8313
EXAMINER
MINCEY, JERMAINE A
ART UNIT PAPER NUMBER
2159
NOTIFICATION DATE DELIVERY MODE
06/30/2016 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):
USPTO@wavsip.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
BEFORE THE PATENT TRIAL AND APPEAL BOARD
Exparte ALOK S. KIRPAL
Appeal2013-007782
Application 12/013,289
Technology Center 2100
Before CAROLYN D. THOMAS, DEBRA K. STEPHENS, and
NABEEL U. KHAN, Administrative Patent Judges.
STEPHENS, Administrative Patent Judge.
DECISION ON APPEAL
STATEMENT OF THE CASE
Appellant appeals under 35 U.S.C. § 134 from a Final Rejection of
claims 1, 3-5, 7-10, 12-14, and 16-21. 1 We have jurisdiction under
35 U.S.C. § 6(b).
We AFFIRM.
1 Claims 2, 6, 11, and 15 have been cancelled.
Appeal2013-007782
Application 12/013,289
STATEMENT OF THE fNVENTION
According to Appellant, the claims are directed to extracting entities
from a web page using a high precision low recall technique and a sequential
model (Abstract). Claim 1, reproduced below, is representative of the
claimed subject matter:
1. A method comprising:
training a high precision low recall (HPLR), template-
based technique on a first web page, producing one or more
annotated entities extracted from the first web page so that the
HPLR, template-based technique is trained to extract annotated
entities from a first cluster of web pages that are structurally
similar and for which a template has been created prior to
application of the HPLR technique on such first cluster,
including the first web page;
extracting one or more annotated entities from the first
cluster of web pages using the HPLR, template-based technique;
training a sequential model by using the one or more
annotated entities extracted by the HPLR technique as an
observation sequence input to extract additional annotated
entities from the first cluster and one or more other clusters of
web pages for which a template has not been created, wherein
the first cluster's web pages are structurally different than the one
or more other cluster's web pages; and
applying the trained sequential model on a second web
page from a second cluster of the one or more other clusters,
producing one or more annotated entities extracted from the
second web page.
REFERENCES
The prior art relied upon by the Examiner in rejecting the claims on
appeal is:
Shanahan
Wen
US 2005/0228783 Al
US 2008/0046441 Al
2
Oct. 13, 2005
Feb.21,2008
Appeal2013-007782
Application 12/013,289
V.G. Vinod Vydiswaran & Sunita Sarawagi, Advances in Data Management,
"Learning to extract information from large websites using sequential
models" 3-14 (2005).
Philippe Le Hegaret, The W3C Document Object Model (DOM), 2002,
available at http:www.w3.org/2002/07/26-dom-article (last visited Mar. 16,
2010).
REJECTIONS
Claims 1, 4, 5, 9, 10, 13, 14, and 18-21 stand rejected under 35 U.S.C.
§ 103(a) as being unpatentable over Shanahan and Vinod (Final Act. 3-11).
Claims 3, 7, 12, and 16 stand rejected under 35 U.S.C. § 103(a) as
being unpatentable over Shanahan, Vinod, and Wen (Final Act. 12-13).
Claims 8 and 17 stand rejected under 35 U.S.C. § 103(a) as being
unpatentable over Shanahan, Vinod, Wen, and Hegaret (Final Act. 14--15).
We have only considered those arguments that Appellant actually
raised in the Briefs. Arguments Appellant could have made but chose not to
make in the Briefs have not been considered and are deemed to be waived.
See 37 C.F.R. § 41.37(c)(l)(iv) (2012).
ISSUES
35 US.C. § 103(a): Claims 1, 4, 5, 9, 10, 13, 14, and 18-21
Appellant asserts the invention is not obvious over Shanahan and
Vinod (App. Br. 4--8). The issue presented by the arguments is:
Issue 1: Has the Examiner erred in finding the combination of
Shanahan and Vinod teaches or suggests
training a high precision low recall (HPLR), template-based
technique and
3
Appeal2013-007782
Application 12/013,289
training a sequential model by using the one or more annotated
entities extracted by the HPLR technique as an observation
sequence input to extract additional annotated entities from the
first cluster and one or more other clusters of web pages for
which a template has not been created, wherein the first
cluster's web pages are structurally different than the one or
more other cluster's web pages,
as recited in claim 1, and similarly recited in claims 10 and 19?
Issue 2: Has the Examiner improperly combined the teachings and
suggestions of Shanahan and Vinod?
ANALYSIS
We disagree with Appellant's conclusions and adopt as our own:
( 1) the findings and reasons set forth by the Examiner in the action from
which this appeal is taken; and (2) the reasons set forth by the Examiner in
the i~JlS\ver in response to the i\.ppeal Brief. \Vith respect to the claims
argued by Appellant, we highlight and address specific findings and
arguments for emphasis as follows.
Appellant first argues the combination of Shanahan and Vinod does
not teach or suggest "training a sequential model by using the one or more
annotated entities extracted by the HPLR technique as an observation
sequence input" (App. Br. 4). Specifically, Appellant argues that
Shanahan's Support Vector Machine (SVM) "does not indicate that these
features are extracted 'using [an] HPLR, template-based technique' as
recited in claim 1" (id. at 5). Appellant further argues training of
Shanahan' s SVM is to "satisfy an information need," which requires "a
highly heterogeneous set of documents"; and argues the information
4
Appeal2013-007782
Application 12/013,289
retrieved by the searches in Shanahan "is extracted from documents without
the use of a template to determine whether the document 'satisfies the
information need' already expressed" in contrast to Appellant's recited
invention (id. at 5---6).
We are not persuaded by Appellant's arguments. Initially, we note
"HPLR," "HPLR technique," and "HPLR, template-based technique" are not
defined explicitly in Appellant's Specification. Appellant points to
paragraph three of the Specification as describing a "template-based
technique" which states "[a] class of techniques utilized to extract entities
and attributes from web pages is known as template-based techniques"
(Reply Br. 9 (citing Spec. i-f 3)).
The Examiner finds, and we agree, Shanahan discloses acquiring an
SVM through a process of learning where "[t]he learnt model tends to
exhibit[] high precision, but very low recall" (Shanahan i-f 16; Ans. 21 ). We
further agree with the Examiner's finding that the SVM disclosed in
Shanahan is a high precision low recall technique that models a hyperplane
of classified objects, and that Shanahan teaches the documented information
form complex constructs (Ans. 21-22; Shanahan i-f 35). We agree with the
Examiner that the support vector machines serve as a computational model
or template (Ans. 21 ).
We further agree with the Examiner that the structural features
described in Shanahan are annotated, or marked, entities (Ans. 22); and
neither Appellant's Specification nor the claim itself expressly defines what
an annotated entity is. Rather, Appellant's Specification merely describes
examples of annotated entities (see Cl. 1; see also Spec. i-fi-1 3--4, 14--15).
5
Appeal2013-007782
Application 12/013,289
As set forth by the Examiner, Shanahan further discloses many
alternative techniques are known in the art for extracting these structural
features, storing the features, and sorting or ranking features (Ans. 22).
Therefore, we are not persuaded Shanahan fails to teach an HPLR, template-
based technique.
Appellant additionally argues Vinod clearly does not teach or suggest
"training a sequential model by using ... one or more annotated entities
extracted by the HPLR technique as an observation sequence input" (App.
Br. 7). Appellant concedes that "Vinod describes training a sequential
model ... a Conditional Random Field (CRF);" but Appellant argues Vinod
does not teach or suggest the training is based on an HPLR technique (id. at
6-7).
We are not persuaded by Appellant's arguments. Appellant concedes
that "Vinod describes training a sequential model ... a Conditional Random
Field (CRF)" (App. Br. 6); and Appellant's own Specification lists CRF as
an exemplary technique to perform sequential modeling (see Spec. i-f 12).
The Examiner finds, and we agree, that Vinod discloses the training of a
sequential model to extract, in a second phase, additional annotated entities
fetched from relevant web pages (Ans. 23).
Appellant also argues Shanahan and Vinod in combination, cannot
teach or disclose the features of claim 1 (App. Br. 7). Specifically,
Appellant argues (1) Shanahan does not teach a template-based technique
and (2) the training technique disclosed in Vinod is based on selected
hyperlinks, not structurally similar websites (id.). Therefore, Appellant's
argue, "even if the teachings of Shanahan could properly be characterized as
describing a template-based technique ... one of ordinary skill in the art
6
Appeal2013-007782
Application 12/013,289
would not be motivated to use such a technique in the context of Vinod in
the manner suggested by the Examiner" (id. at 8).
We are not persuaded by Appellant's arguments. Regarding the
template based technique disclosed in Shanahan, we agree with the
Examiner's findings as discussed above. The Examiner finds, and we agree,
that Vinod teaches an HPLR technique in the feature extraction during the
page fetch, having a structure format and parsing text content (Ans. 23).
Therefore, we agree with the Examiner's finding that the combination of
Shanahan and Vinod would have been obvious to one of ordinary skill in the
art (Final Act. 5; Ans. 23).
We are not persuaded by Appellant that an ordinarily skilled artisan
would not have been motivated to combine the teachings and suggestions of
Shanahan (App. Br. 8). The Examiner has articulated reasoning with some
rational underpinning - to train a model to automatically find desired pages
- citing Vydiswaran's Introduction. Appellant has not proffered sufficient
evidence or argument to persuade us of Examiner error.
Accordingly, we are not persuaded the Examiner erred in finding the
combination of Shanahan and Vinod teaches or suggests the limitations as
recited in claim 1 and claims 4, 5, 9, 10, 13, 14, and 18-21, not separately
argued. Therefore, we sustain the rejection of claims 1, 4, 5, 9, 10, 13, 14,
and 18-21under35 U.S.C. § 103(a) for obviousness over Shanahan and
Vinod.
35 US.C. § 103(a): Claims 3, 7, 8, 12, 16, and 17
Appellant asserts the invention as recited in claims 3, 7, 12, and 16 is
not unpatentable over the combination of Shanahan, Vinod, and Wen; and
7
Appeal2013-007782
Application 12/013,289
the invention as recited in claims 8 and 17 is not unpatentable over the
combination of Shanahan, Vinod, Wen, and Hegaret, respectively, relying
on the arguments set forth for claim 1 (App. Br. 8). For the reasons set forth
above, we are not persuaded of error in the Examiner's findings and
conclusion. Accordingly, we sustain the rejection of claims 3, 7, 12, 16, and
8 and 17 under 35 U.S.C. § 103(a) for obviousness over Shanahan, Vinod,
and Wen; and Shanahan, Vinod, Wen, and Hegaret, respectively.
DECISION
The Examiner's rejection of claims 1, 4, 5, 9, 10, 13, 14, and 18-21
under 35 U.S.C. § 103(a) as being unpatentable over Shanahan and Vinod is
AFFIRMED.
The Examiner's rejection of claims 3, 7, 12, and 16 under 35 U.S.C.
§ 103(a) as being unpatentable over Shanahan, Vinod, and Wen is
AFFIRMED.
The Examiner's rejection of claims 8 and 17 under 35 U.S.C. § 103(a)
as being unpatentable over Shanahan, Vinod, Wen, and Hegaret is
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). See 37 C.F.R.
§ 1.136(a)(l )(iv).
AFFIRMED
8