Ex Parte Bharat et alDownload PDFPatent Trial and Appeal BoardNov 12, 201310750363 (P.T.A.B. Nov. 12, 2013) Copy Citation 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. 10/750,363 12/31/2003 Krishna Bharat Google-44 (GP-096-00-US) 4908 82402 7590 11/12/2013 Straub & Pokotylo 788 Shrewsbury Avenue Tinton Falls, NJ 07724 EXAMINER AUGUSTIN, EVENS J ART UNIT PAPER NUMBER 3685 MAIL DATE DELIVERY MODE 11/12/2013 PAPERPAPER 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 1 ___________ 2 3 BEFORE THE PATENT TRIAL AND APPEAL BOARD 4 ___________ 5 6 Ex parte KRISHNA BHARAT, STEPHEN LAWRENCE 7 and MEHRAN SAHAMI 8 ___________ 9 10 Appeal 2011-008040 11 Application 10/750,363 12 Technology Center 3600 13 ___________ 14 15 16 Before ANTON W. FETTING, JOSEPH A. FISCHETTI, and 17 BIBHU R. MOHANTY, Administrative Patent Judges. 18 FETTING, Administrative Patent Judge. 19 DECISION ON APPEAL 20 21 Appeal 2011-008040 Application 10/750,363 2 STATEMENT OF THE CASE1 1 Krishna Bharat, Stephen Lawrence and Mehran Sahami (Appellants) 2 seek review under 35 U.S.C. § 134 of a final rejection of claims 1, 3, 5-26, 3 33, 35, 37-58, 65 and 66, the only claims pending in the application on 4 appeal. We have jurisdiction over the appeal pursuant to 35 U.S.C. § 6(b). 5 Appellants invented a way of determining particularly relevant 6 advertisements or advertisement creatives to serve for a user request, such as 7 a search query or document request for example (Specification 1:7-10). 8 An understanding of the invention can be derived from a reading of 9 exemplary claim 1, which is reproduced below [bracketed matter and some 10 paragraphing added]. 11 1. A computer-implemented method for determining user 12 profile information for a user, the computer-implemented 13 method comprising: 14 a) determining, with a computer system including at least one 15 computer on a network, 16 initial user profile information 17 for the user 18 using information included in past search queries 19 submitted to a search engine by the user, 20 wherein such information is independent of documents 21 returned as search results to the past search 22 queries; 23 b) inferring, with the computer system, 24 user profile information for the user; 25 1 Our decision will make reference to the Appellants’ Appeal Brief (“App. Br.,” filed August 12, 2010) and the Examiner’s Answer (“Ans.,” mailed January 6, 2011). Appeal 2011-008040 Application 10/750,363 3 c) determining, with the computer system, the user profile 1 information 2 for the user 3 using both 4 the initial user profile information 5 and 6 the inferred user profile information; 7 and 8 d) controlling, with the computer system, the serving 9 of an advertisement 10 to the user 11 using the determined user profile information. 12 The Examiner relies upon the following prior art: 13 Rose US 5,724,567 Mar. 3, 1998 Herz US 5,754,939 May 19, 1998 Claims 1, 3, 5-26, 33, 35, 37-58, 65 and 66 stand rejected under 35 14 U.S.C. § 103(a) as unpatentable over Herz and Rose. 15 ISSUES 16 The issues of obviousness turn primarily on the breadth and degree of 17 patentable weight afforded the term “user profile information,” and whether 18 plural similar queries were predictable. 19 FACTS PERTINENT TO THE ISSUES 20 The following enumerated Findings of Fact (FF) are believed to be 21 supported by a preponderance of the evidence. 22 Appeal 2011-008040 Application 10/750,363 4 Facts Related to the Prior Art 1 Herz 2 01. Herz is directed to customized electronic identification of 3 desirable objects, such as news articles, in an electronic media 4 environment, and in particular to automatically constructing both a 5 “target profile” for each target object in the electronic media 6 based, for example, on the frequency with which each word 7 appears in an article relative to its overall frequency of use in all 8 articles, as well as a “target profile interest summary” for each 9 user, which target profile interest summary describes the user’s 10 interest level in various types of target objects. The system then 11 evaluates the target profiles against the users’ target profile 12 interest summaries to generate a user-customized rank ordered 13 listing of target objects most likely to be of interest to each user so 14 that the user can select from among these potentially relevant 15 target objects, which were automatically selected by this system 16 from the plethora of target objects that are profiled on the 17 electronic media. Herz 1:18-34. 18 02. Relevant definitions of terms for the purpose of following Herz’s 19 description include: (a.) an object available for access by the user, 20 which may be either physical or electronic in nature, is termed a 21 “target object,” (b.) a digitally represented profile indicating that 22 target object’s attributes is termed a “target profile,” (c.) the user 23 looking for the target object is termed a “user,” (d.) a profile 24 holding that user’s attributes, including age/zip code/etc. is termed 25 a “user profile,” (e.) a summary of digital profiles of target objects 26 Appeal 2011-008040 Application 10/750,363 5 that a user likes and/or dislikes, is termed the “target profile 1 interest summary” of that user, (f) a profile consisting of a 2 collection of attributes, such that a user likes target objects whose 3 profiles are similar to this collection of attributes, is termed a 4 “search profile” or in some contexts a “query” or “query profile,” 5 (g.) a specific embodiment of the target profile interest summary 6 which comprises a set of search profiles is termed the “search 7 profile set” of a user, (h.) a collection of target objects with similar 8 profiles, is termed a “cluster,” (i.) an aggregate profile formed by 9 averaging the attributes of all target objects in a cluster, termed a 10 “cluster profile,” (j.) a real number determined by calculating the 11 statistical variance of the profiles of all target objects in a cluster, 12 is termed a “cluster variance,” (k.) a real number determined by 13 calculating the maximum distance between the profiles of any two 14 target objects in a cluster, is termed a “cluster diameter.” Herz 15 4:48 – 5:5. 16 03. Each movie has a different set of values for its attributes. 17 Attributes c-g are numeric attributes, of the sort that might be 18 found in a database record. It is evident that they can be used to 19 help the user identify target objects (movies) of interest. For 20 example, the user might previously have rented many Parental 21 Guidance (PG) films, and many films made in the 1970’s. This 22 generalization is useful: new films with values for one or both 23 attributes that are numerically similar to these (such as MPAA 24 rating of 1, release date of 1975) are judged similar to the films 25 the user already likes, and therefore of probable interest. 26 Appeal 2011-008040 Application 10/750,363 6 Attributes a-b and h are textual attributes. They too are important 1 for helping the user locate desired films. For example, perhaps the 2 user has shown a past interest in films whose review text (attribute 3 h) contains words like “chase,” “explosion,” “explosions,” “hero,” 4 “gripping,” and “superb.” This generalization is again useful in 5 identifying new films of interest. Attribute i is an associative 6 attribute. It records associations between the target objects in this 7 domain, namely movies, and ancillary target objects of an entirely 8 different sort, namely humans. A good indication that the user 9 wants to rent a particular movie is that the user has previously 10 rented other movies with similar attribute values, and this holds 11 for attribute i just as it does for attributes a-h. For example, if the 12 user has often liked movies that customer C17 and customer C190 13 have rented, then the user may like other such movies, which have 14 similar values for attribute i. Attribute j is another example of an 15 associative attribute, recording associations between target objects 16 and actors. Notice that any of these attributes can be made subject 17 to authentication when the profile is constructed, through the use 18 of digital signatures; for example, the target object could be 19 accompanied by a digitally signed note from the MPAA, which 20 note names the target object and specifies its authentic value for 21 attribute c. Herz 10:25-60. 22 04. The simplest way to use the automatic menuing system is for the 23 user to begin browsing at the top of the tree and moving to more 24 specific subclusters. The user may optionally provide a query 25 consisting of textual and/or other attributes, from which query the 26 Appeal 2011-008040 Application 10/750,363 7 system constructs a profile, optionally altering textual attributes 1 before decomposing them into numeric attributes. Query profiles 2 are similar to the search profiles in a user’s search profile set, 3 except that their attributes are explicitly specified by a user, most 4 often for one-time usage, and unlike search profiles, they are not 5 automatically updated to reflect changing interests. The system 6 automatically locates a small set of one or more clusters with 7 profiles similar to the query profile. The user may start browsing 8 at any of these clusters, and can move from it to subclusters, 9 superclusters, and other nearby clusters. For a user who is looking 10 for something in particular, it is generally less efficient to start at 11 the largest cluster and repeatedly select smaller subclusters than it 12 is to write a brief description of what one is looking for and then 13 to move to nearby clusters if the objects initially recommended are 14 not precisely those desired. Herz 66:45 – 67:7. 15 05. In a system where queries are used, it is useful to include in the 16 target profiles an associative attribute that records the associations 17 between a target object and whatever terms are employed in 18 queries used to find that target object. The association score of 19 target object X with a particular query term T is defined to be the 20 mean relevance feedback on target object X, averaged over just 21 those accesses of target object X that were made while a query 22 containing term T was active, multiplied by the negated logarithm 23 of term T’s global frequency in all queries. The effect of this 24 associative attribute is to increase the measured similarity of two 25 documents if they are good responses to queries that contain the 26 Appeal 2011-008040 Application 10/750,363 8 same terms. A further maneuver can be used to improve the 1 accuracy of responses to a query: in the summation used to 2 determine the quality q(U, X) of a target object X, a term is 3 included that is proportional to the sum of association scores 4 between target object X and each term in the active query, if any, 5 so that target objects that are closely associated with terms in an 6 active query are determined to have higher quality and therefore 7 higher interest for the user. To complement the system’s 8 automatic reorganization of the hierarchical cluster tree, the user 9 can be given the ability to reorganize the tree manually, as he or 10 she sees fit. Any changes are optionally saved on the user’s local 11 storage device so that they will affect the presentation of the tree 12 in future sessions. Herz 69:45 – 70:3. 13 Rose 14 06. Rose is directed to information access in multiuser computer 15 systems, and more particularly to a computer-based information 16 system that enables users to access information from a wide 17 variety of sources. Rose 1:6-9. 18 ANALYSIS 19 Claims 1, 3, 5, 6, 33, 35, 37 and 38 20 We are not persuaded by Appellants’ argument that 21 the cited patents do net teach or make obvious determining ... 22 initial user profile information for the user using information 23 included in past search queries submitted to a search engine 24 by the user, wherein such information is independent of 25 documents returned as search results to the past search 26 queries. 27 Appeal 2011-008040 Application 10/750,363 9 The Examiner contends that the “query profile” discussed 1 on column 66, lines 57-67 of the Herz patent teaches this 2 feature. [] The appellant respectfully disagrees. More 3 specifically, the “query profile” discussed on column 66, lines 4 57-67 of the Herz patent includes information from a single 5 search query (as opposed to plural search queries as claimed), 6 and can be updated using document cluster information (as 7 opposed to being independent of the documents returned as 8 search results as claimed). 9 App. Br. 14-15. Claim 1 recites “determining [] initial user profile 10 information for the user using information included in past search queries.” 11 The step merely requires determining some user profile information. This 12 determination is further recited as “using” in some unspecified manner 13 information included in past search queries. The step does not extract 14 information included in past search queries or even perform those queries. 15 Thus, the step relies on some arbitrary bit pattern that in the user’s mind 16 stands for information included in past search queries. On that basis alone, 17 we would afford the nature of such information no patentable weight. See 18 In re Ngai, 367 F.3d 1336, 1339 (Fed. Cir. 2004) and King Pharms, Inc. v. 19 Eon Labs, Inc., 616 F.3d 1267, 1279 (Fed. Cir. 2010). 20 However, the Examiner found the information in Herz used information 21 included in a past search query. This is uncontested. The argument is this 22 does not show plural such queries. As this is a rejection under obviousness, 23 it was at least predictable that a user would perform plural similar queries in 24 the past, as searching is an inexact science and a first query frequently 25 misses the mark. 26 Claims 14-19, 46-51, 65, and 66 27 We are not persuaded by Appellants’ argument that 28 Appeal 2011-008040 Application 10/750,363 10 [a]lthough movies can have attributes including a “list of 1 customers who have previously rented this movie” (See, e.g., 2 column 10, lines 22 and 23), such a list, by itself, is not user 3 profile information for a document, which is associated with 4 the document, and which association is stored, as recited in 5 claims 14 and 46. That is, a list of customers, by itself, is 6 insufficient to provide attributes of those customers. 7 App. Br. 27-28. Claim 14 recites “determining [] initial user profile 8 information for the document” and “associating [] with the document, the 9 determined user profile information for the document.” The steps merely 10 require determining some user profile information that is in any manner for a 11 document and making some form of association between the document and 12 the determined user profile information. 13 The steps do not narrow the relationships between the profile and the 14 document. In fact the determination itself is a form of association. The 15 steps do not preclude implicit association, so long as that association can in 16 some sense be stored with the user profile information. 17 The Examiner found 18 [t]he system can relate a user with past searches words 19 such past interest in films whose review text (attribute h) 20 contains words like “chase,” “explosion,” “explosions,” “hero,” 21 “gripping,” and “superb” (column 10, lines 37-42). The system 22 can also record associations between documents (movies) and 23 users column 10, lines 43-46). A good indication that the user 24 wants to rent a particular movie is that the user has previously 25 rented other movies with similar attribute values. For example, 26 if the user has often liked movies that customer 1 and customer 27 2 have rented, then the user may like other such movies. Since 28 the system can system relationships between users and 29 documents one skilled in the art could easily infer from these 30 relationships to create graphs (column 10, lines 46-53). 31 Appeal 2011-008040 Application 10/750,363 11 Final Rej. 5-6. Again, Appellants’ arguments are not commensurate with 1 the scope of the claim, which does not narrow the content or character of 2 user profile information. Appellants simply declare Herz’s information is 3 not user profile information without showing how Herz’s information 4 violates any lexicographic definition of such information. Also, as with 5 claim 1, the characteristic of data being user profile information is solely 6 interpretable by the human mind and as such is deserving of no patentable 7 weight. 8 Claims 7, 13, 20, 39, 45 and 52 9 We are not persuaded by Appellants’ argument that Herz’s 10 nodes and links are in no way related to nodes and edges of a 11 graph, the topology of which is used to infer user profile 12 information. 13 Appeal Br. 16. Claim 7 recites 14 i) defining a node for each of a number of documents and the 15 user, wherein each node represents a particular one of the 16 number of documents or the user, 17 ii) adding edges between nodes if there is an association 18 between the nodes to define a graph, wherein there is an 19 association between at least two of the nodes, and 20 iii) inferring user profile information for the user using a 21 topology of the graph and user profile information of other 22 documents. 23 The steps merely require defining nodes and edges and making 24 inferences. Nodes and edges are abstract data types, implementable by any 25 number of techniques. No implementation is recited or defined. Inference 26 may be in the mind of the beholder. 27 Herz shows several examples of nodes and edges where the nodes are 28 documents. Herz figs. 6-9. A tree as used in Herz is mathematically simply 29 Appeal 2011-008040 Application 10/750,363 12 an acyclic complete graph of nodes and edges. Each edge in a tree defines 1 an association between its endpoint nodes. Thus these figures show 2 elements (i) and (ii) of nodes and edges. As the claim does not narrow the 3 content of the user profile information that is inferred, the very tree itself is 4 part of the profile of the user creating it and so presents such an inference. 5 Claims 8-12, 21-26, 40-44, and 53-58 6 Claims 8-12, 21-26, 40-44, and 53-58 each recite either an implicative 7 step that is triggered or a multiplicative step applied to neighboring nodes. 8 The Examiner found the implicative steps did not need to be performed. 9 This is erroneous because, as Appellants argue, each of these claims triggers 10 the implicative step. The Examiner made no findings as to the multiplicative 11 steps being applied to neighboring nodes. The Examiner thus failed to 12 present a prima facie case as to these claims. 13 CONCLUSIONS OF LAW 14 The rejection of claims 1, 3, 5, 6, 7, 13, 14-19, 20, 33, 35, 37, 38, 39, 45, 15 46-51, 52, 65, and 66 under 35 U.S.C. § 103(a) as unpatentable over Herz 16 and Rose is proper. 17 The rejection of claims 8-12, 21-26, 40-44, and 53-58 under 35 U.S.C. 18 § 103(a) as unpatentable over Herz and Rose is improper. 19 DECISION 20 The rejection of claims 1, 3, 5, 6, 7, 13, 14-19, 20, 33, 35, 37, 38, 39, 45, 21 46-51, 52, 65, and 66 is affirmed. 22 The rejection of claims 8-12, 21-26, 40-44, and 53-58 is reversed. 23 Appeal 2011-008040 Application 10/750,363 13 No time period for taking any subsequent action in connection with this 1 appeal may be extended under 37 C.F.R. § 1.136(a). See 37 C.F.R. 2 § 1.136(a)(1)(iv) (2011). 3 4 AFFIRMED-IN-PART 5 6 7 8 9 10 11 mls 12 Copy with citationCopy as parenthetical citation