Ex Parte Acharya

Board of Patent Appeals and Interferences

Jun 18, 2012

11284603 (B.P.A.I. Jun. 18, 2012)

UNITED STATES PATENT AND TRADEMARKOFFICE
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.
11/284,603 11/21/2005 Chiranjit Acharya 7114-86640-US 6671
37123 7590 06/19/2012
FITCH EVEN TABIN & FLANNERY, LLP
120 SOUTH LASALLE STREET
SUITE 1600
CHICAGO, IL 60603-3406
EXAMINER
RICHARDSON, JAMES E
ART UNIT PAPER NUMBER
2167
MAIL DATE DELIVERY MODE
06/19/2012 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 BOARD OF PATENT APPEALS
AND INTERFERENCES
________________

Ex parte CHIRANJIT ACHARYA
________________

Appeal 2010-003919
Application 11/284,603
Technology Center 2100
________________

Before ROBERT E. NAPPI, KRISTEN L. DROESCH, and JOHN G.
NEW, Administrative Patent Judges.

NEW, Administrative Patent Judge.

DECISION ON APPEAL

Appellant appeals under 35 U.S.C. § 134(a) from the Examiner’s
rejection of claims 1-18, which stand rejected under 35 U.S.C. § 103(a) as
being unpatentable over U.S. Patent Publication No. 2004/0054572 A1 to
Oldale et al. (“Oldale”), in view of U.S. Patent No. 6,981,040 B1 to Konig
et al. (“Konig”), and also in view of Lyle H. Ungar, et al., A Formal
Statistical Approach to Collaborative Filtering, Conference on Automated
Learning and Discovery, 1-6 (1998) (“Ungar”).
Appeal 2010-003919
Application 11/284,603

2
We reverse.
STATEMENT OF THE CASE
Appellant describes the present invention, entitled User's Preference
Prediction from Collective Rating Data as follows:
A computer-implemented method includes receiving a dataset
representing a plurality of users, a plurality of items, and a
plurality of ratings given to items by users; clustering the
plurality of users into a plurality of user-groups such that at
least one user belongs to more than one user-group; clustering
the plurality of items into a plurality of item-groups such that at
least one item belongs to more than one item-group; inducing a
model describing a probabilistic relationship between the
plurality of users, items, ratings, user-groups, and item-groups,
the induced model defined by a plurality of model parameters;
and predicting a rating of a user for an item using the induced
model.

Abstract.

Independent claim 1 is representative1:

A computer-implemented method, comprising:

obtaining a dataset representing a plurality of users, a plurality
of items, and a plurality of ratings given to items by users;

clustering the plurality of users into a plurality of user-groups
such that at least one user belongs to more than one user-group;

clustering the plurality of items into a plurality of item-groups
such that at least one item belongs to more than one item-group;

1 Appellant and Examiner agree that the Examiner’s rejection of
independent claims 1 and 10 were based upon the same reasoning.
Appellant’s Brief (App. Br.) 18; Examiner’s Answer (Ex. Ans.) 4-8 and 12-
15. Consequently, we choose claim 1 as representative.
Appeal 2010-003919
Application 11/284,603

3
inducing a model describing a probabilistic relationship
between the plurality of users, items, ratings, user-groups, and
item-groups, the induced model defined by a plurality of model
parameters; and

predicting a rating of a user for an item using the induced
model.

Claims 2-9 depend from claim 1 and claims 11-18 depend from claim
10. Appellant admits that, for purposes of the instant appeal, the applicant
is content to rely upon the arguments raised with respect to claims 1 and 10
for all of the claims.
ISSUES
Claims 1 and 10
The Examiner concludes that the claims are unpatentable as obvious
under 35 U.S.C. § 103(a) over the combination of prior art references
Oldale, Konig, and Ungar. Specifically, the Examiner concludes that it
would have been obvious for an artisan of ordinary skill to combine the
teachings of Oldale with the teachings of Konig by modifying Oldale such
that when customers of Oldale are sorted into groups or clusters based on
profile similarity, a user is sorted into multiple clusters based on similarities
to multiple groups as in Konig. Ex. Ans. 6.
Furthermore, the Examiner finds, although neither Oldale nor Konig
specifically disclose inducing a model describing a probabilistic
relationship between the plurality of user-groups, and item-groups, Ungar
discloses inducing a model describing a probabilistic relationship between a
plurality of user-groups and item-groups. Ex. Ans. 7. The Examiner
concludes, at the time of invention it would have been obvious to a person
having ordinary skill in the art to combine the teachings of Oldale and
Appeal 2010-003919
Application 11/284,603

4
Konig with the teachings of Ungar. Ex. Ans. 7. The motivation for so
doing would have been to allow the combined system of Oldale and Konig
to include a probabilistic model in which there are link probabilities
between clusters of users and items. Ex. Ans. 8. Did the Examiner err in
concluding that it would have been obvious to a person of ordinary skill in
the art to combine the teachings of Oldale, Konig, and Ungar, thereby
rendering Appellant’s claimed invention obvious at the time of invention?
ANALYSIS
For the Examiner to establish a prima facie case of obviousness in
view of a combination of prior art references, a proper analysis under § 103
requires, inter alia, consideration of two factors: (1) whether the prior art
would have suggested to those of ordinary skill in the art that they should
make the claimed composition or device, or carry out the claimed process;
and (2) whether the prior art would also have revealed that in so making or
carrying out, those of ordinary skill would have a reasonable expectation of
success. See In re Dow Chemical Co., 837 F.2d 469, 473 (Fed. Cir. 1988).
Because the Examiner has failed to meet at least one of these requirements,
we reverse the Examiner’s rejection of the claims.
Claims 1 and 10 both recite “inducing a model describing a
probabilistic relationship between the plurality of users, items, ratings, user-
groups, and item-groups, the induced model defined by a plurality of model
parameters.” The Examiner finds that Ungar discloses “inducing a model
describing a probabilistic relationship between the plurality of user-groups,
and item-groups.” Ex. Ans. 7. The Examiner points to Ungar’s teaching of
Gibbs Sampling as a “‘probabilistic model in which people and the items
they view or buy are each divided into (unknown) clusters and there are link
Appeal 2010-003919
Application 11/284,603

5
probabilities between these clusters.’” Id. The Examiner further contends
that, at the time of invention, it would have been obvious for a person of
ordinary skill in the art to combine the teachings of Oldale and Konig and
to incorporate the Gibbs sampling taught by Ungar into the model
describing a probabilistic relationship between the plurality of users, items,
and ratings, the induced model defined by:
[A] plurality of model parameters (Oldale, [0017], Lines 8-12)
of Oldale to allow the system to include a probabilistic model in
which there are link probabilities between clusters of users and
items (Ungar, §6 “Summary”, Lines 1-3), and thus allowing the
system to induce a model describing a probabilistic relationship
between the plurality of users, items, ratings, user-groups, and
item-groups. (Ungar, §6 “Summary”, Lines 1-3).

Id. 7-8. The examiner concludes that “[t]he motivation for doing so would
have been to allow the combined system of Oldale and Konig to include a
probabilistic model in which there are link probabilities between clusters of
users and items.” Id. 8.
The Appellant argues it would not have been obvious to an artisan of
ordinary skill to rely upon the probabilistic Gibbs sampling method taught
in Ungar, because Ungar teaches away from the method disclosed in
Appellant’s claims 1 and 10. App. Br. 15. Those claims require, in
relevant part: “clustering the plurality of users into a plurality of user-
groups such that at least one user belongs to more than one user-group” and
also “clustering the plurality of items into a plurality of item-groups such
that at least one item belongs to more than one item-group.” App. Br. 24.
Appellant points out that the “Gibbs sampling method disclosed by Ungar
uses model estimation methods that require random assignment of users or
Appeal 2010-003919
Application 11/284,603

6
items to random groups and using those groups to generate grouping of the
associated users or items.” App. Br. 14. Furthermore, Appellant notes that
Ungar teaches that this method requires that users and items are each
assigned to an individual user- or item-group. Id. (citing Ungar p.3, fn1).
We find Appellant’s reasoning persuasive. Ungar teaches Gibbs
sampling as a method of inducing a probabilistic model relating users to
items. Ungar at 3. Gibbs sampling alternates between two steps: (1)
Assignment, in which a user or item is chosen at random and assigned it a
user- or item-group proportionally to probability of the user- or item-group
generating it and (2) Model estimation, in which one picks the probabilities
that a (random) user is in that given user-group, that a random item is in
that given item-group, and the link probability that a user in a given user-
group is linked to an item in a given item-group. Id.
However, the model taught by Ungar requires that a user or an item
be assigned randomly to a single user- or item-group. App. Br. 14.
Assigning an individual user or an item to more than one group destroys the
functioning of the model. App. Br. 15; see also Ungar 4 (“a person or
movie is picked at random, and then assigned to a class”). The assignment
of a user or item to a given user-group or item-group is fundamentally
incompatible with the requirements of claims 1 and 10 that at least one
person and one item be assigned to more than one user-group or item-group
respectively. Thus, we find that the skilled artisan would be discouraged
from using a probabilistic model in which there are link probabilities
between clusters of users and items and as such we find Ungar to teach
away from the combining the teachings as set forth in the Examiner’s
answer.
Appeal 2010-003919
Application 11/284,603

7
Because Ungar teaches away from the methods disclosed by
Appellant’s independent claims 1 and 10, we disagree with the Examiner
that a person of ordinary skill in the art would combine Ungar with the
teachings of Oldale and Konig. Consequently we reverse the Examiner’s
rejection of claims 1, 10 and the claims which depend thereupon.
Appellant has advanced other arguments with respect to the
obviousness of the disputed claims in light of Oldale and Konig, but we
need not reach those arguments because, for the reasons stated above, we
find the Examiner’s erroneous rejection of the claims based on Ungar to be
dispositive of the case.
CONCLUSION
Appellant has shown that the Examiner erred in rejecting claims 1
through 18 under §103(a).
DECISION

The Examiner’s decision rejecting claims 1-18 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). See 37 C.F.R.
§ 1.136(a)(1)(iv) (2010).
REVERSED

tsj