Shanchan Wu et al.Download PDFPatent Trials and Appeals BoardMay 20, 202014893606 - (D) (P.T.A.B. May. 20, 2020) 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. 14/893,606 11/24/2015 Shanchan Wu 84387709 3549 22879 7590 05/20/2020 HP Inc. 3390 E. Harmony Road Mail Stop 35 FORT COLLINS, CO 80528-9544 EXAMINER BURKE, TIONNA M ART UNIT PAPER NUMBER 2176 NOTIFICATION DATE DELIVERY MODE 05/20/2020 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): ipa.mail@hp.com jessica.pazdan@hp.com yvonne.bailey@hp.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE BEFORE THE PATENT TRIAL AND APPEAL BOARD Ex parte SHANCHAN WU and JERRY LIU Appeal 2018-009059 Application 14/893,606 Technology Center 2100 Before DAVID C. MCKONE, JOHN P. PINKERTON, and CARL L. SILVERMAN, Administrative Patent Judges. SILVERMAN, Administrative Patent Judge. DECISION ON APPEAL STATEMENT OF THE CASE Pursuant to 35 U.S.C. § 134(a), Appellant1 appeals from the Examiner’s decision to reject claims 1–20, which constitute all pending claims. See Final Act. 1. We have jurisdiction under 35 U.S.C. § 6(b). We Reverse. 1 We use the word “Appellant” to refer to “applicant” as defined in 37 C.F.R. § 1.42. Appellant identifies the real party in interest as Hewlett- Packard Development Company, LP. Appeal Br. 4. Appeal 2018-009059 Application 14/893,606 2 CLAIMED SUBJECT MATTER The invention relates to web page output selection. Abstract; Spec. ¶¶ 7–9, Fig. 2. Claim 1, reproduced below, is illustrative of the claimed subject matter (emphases added): 1. A computing system, comprising: a storage to store information about previously output selections of web pages with different sections, wherein outputting selections comprises at least one of printing, digitally clipping, transmitting, and storing; and a processor to: compare a set of features to a content and style of the different sections of a web page to determine features within the different sections; assign feature values to each of the set of features of the different sections; weight, with a set of weights determined by a logistic regression machine learning method using the information, the features of the output selections according to the frequency of the previously output selections sections including the feature; assign a score to different sections of a web page tree where the score is based on the feature values based on the features present in the different sections and the respective weight of the set of weights of the present features; determine sections to output based on the scores of the different sections; store the information in the storage based on the determined sections to output to allow the set of weights to be updated; and cause the determined sections to be output. Appeal Br. 19 (Claim Appendix). Appeal 2018-009059 Application 14/893,606 3 REFERENCES The prior art relied upon by the Examiner is: Name Reference Date Lim US 2013/0275577 Al Oct. 17, 2013 Boal US 2014/0180811 A1 June 19, 2014 Chenthamarakshan et al. US 2011/0055285 A1 Mar. 3, 2011 Zeng et al. US 2012/0297025 A1 Nov. 22, 2012 Rios et al. US 2002/0194162 A1 Dec. 19, 2002 REJECTIONS Claims 1, 2, 5–12, and 14–19 are rejected under 35 U.S.C. § 103(a) as being unpatentable over Lim in view of Boal. Final Act. 2–16. Claims 3 and 4 are rejected under 35 U.S.C. § 103(a) as being unpatentable over Lim, Boal, and Chenthamarakshan. Id. at 16–18. Claim 13 is rejected under 35 U.S.C. § 103(a) as being unpatentable over Lim, Boal, and Zeng. Id. at 18–19. Claim 20 is rejected under 35 U.S.C. § 103(a) as being unpatentable over Lim, Boal, and Rios. Id. at 19. ANALYSIS Obviousness Rejections Under § 103(a) Claims 1, 2, 5–12, 14–19 over Lim and Boal In the Final Action, the Examiner finds Lim teaches the limitations of claim 1, and independent claims 6 and 11, except Lim does not fully disclose assigning values to, and weighting, features to determine which content to output. Final Act. 4. In particular, the Examiner finds that Lim does not Appeal 2018-009059 Application 14/893,606 4 teach the following claim 1 limitations (also referred to as “disputed limitations”) and relies on Boal for teaching: (L1) assign feature values to each of the set of features of the different sections (The Examiner refers to Boal ¶¶ 168, 169); (L2) weight, with a set of weights determined by a logistic regression machine learning method using the information. (The Examiner refers to the learning component 1774 of Boal using the information and the scoring function to properly weight features of the context. Id. at 4 (citing Boal ¶ 168; Fig. 17); and (L3) store the information in the storage based on the determined sections to output to allow the set of weights to be updated (The Examiner refers to Boal ¶¶ 276, 277). Id. at 4. The Examiner determines it would have been obvious to one having ordinary skill in the art at the time the invention was filed to include weighting features to determine content to be output for the purpose of enhancing a user’s experience by recommending content to be output, as taught by Boal. Id. at 4–5. Appellant argues, inter alia, Boal does not teach disputed limitation L1 because Boal’s offer recommendation uses contextual signals which are not “features within the different sections of a webpage” as in claim 1, and, regarding limitation L2, Boal uses these contextual signals to create scoring functions using a learning component. Appeal Br. 9 (citing Boal ¶¶ 167– 169). According to Appellant these scoring functions are not “a set of weights determined by a logistic regression machine learning method using the information” “about previously output sections of web pages with different sections” Appeal 2018-009059 Application 14/893,606 5 as in claim 1. In fact, Boal uses the association scores with a set of weights to create offer scores. There is no disclosure, teaching, or suggestion in Boal that these weights used with the association scores are determined by a logistic regression machine learning method. Boal only discloses that the scoring functions are determined by a learning component. Further, Boal discloses that the offer scores are used with another set of weights (contextual weighting components) to create universal scores to further create ranked offer sets. These contextual weighting components are also not disclosed, taught, or suggested by Boal as being determined by a logistic regression machine learning method. Id. at 9. In particular, Appellant argues Boal’s learning component learns from objective functions (KPIs) to create scoring functions. Id. at 10–11 (citing Boal ¶ 168; Fig. 17). Appellant further argues Boal’s “contextual weighting 1781 applied to various data [ ] is not taught as being determined by a learning component and certainly not by a logistic regression machine learning method.” Id. at 10–11 (citing Boal, Fig. 17). In the Answer, the Examiner finds Boal teaches executing one or more scoring functions to generate various association scores between context items and offers. These association scores are then used, along with various weights to generate offer scores for each offer. The scores may be weighted by a contextual weighting component and may assign different weights depending on the recommendation context. These scoring functions may have been derived in full or in part manually or may be have been learned by a learning component using any suitable machine learning techniques (see paragraphs [0168][0169]). Although Boal teaches recommending offers, which may be different from the instant application, both use weighted features and scoring functions to recommend an output. Boal uses an association score to associate context items, which is similar to scoring features of content and style, except Boal is scoring context, but both determine items of content to score. Based on the scoring of the context items/features, they are weighted according to the features of Appeal 2018-009059 Application 14/893,606 6 the recommendation context. The association score, along with various weights, generate offer scores assigned to each offer to determine which offers will be recommended and distributed to consumers. Thus, Boal teaches the disclosed limitations. Ans. 20–21. Boal teaches assigning scores to different context and items of an offer to generate offer scores and based on the scores associated with the offer, make a determination whether to recommend the other. Just as in the instant application, features of the offers are scored on how associated they are, then those associations are weighted. Based on the weights, an offer score is generated and ranked. Distributing offers to customers is based on the offer score. Boal teaches a selection process by scoring context, assigning weights to the context and generating scores to determine what to recommend to the user. This application scores features of sections, assigns weights to those features and scores the sections based on the weights, to determine which section to recommend to the user. Id. at 22. In the Reply Brief, Appellant reiterates arguments and argues the Examiner has not responded fully to Appellant’s argument that Boal does not teach weights that are determined by a logistic regression machine learning method. Reply Br. 2–3. In particular, Appellant argues Boal’s use of a learning component based on objective functions (KPIs) to create scoring functions 1775 applied to the various data does not teach “determining a set of weights by a logistic regression machine.” Id. at 3–4. Appellant further argues the contextual weighting 1781 (“a set of weights”) applied to the various data in Fig. 17 is not taught as being determined by a learning component and certainly not by a logistic regression machine learning method. According to Appellant, “the Examiner has not indicated how Boal teaches how the contextual weighting 1781 (‘a set of weights’) is Appeal 2018-009059 Application 14/893,606 7 determined by a logistic machine learning method rather than the scoring functions or that one of skill in the art would be motivated to do so without the claimed invention as a template.” Id. at 3–4. On the record before us, we are persuaded by Appellant’s argument that the Examiner’s finding that the combination of Lim and Boal teaches the disputed limitation L2 is not supported by sufficient evidence or explanation. As stated by the Supreme Court, the Examiner’s obviousness rejection must be based on: [S]ome articulated reasoning with some rational underpinning to support the legal conclusion of obviousness. . . . [H]owever, the analysis need not seek out precise teachings directed to the specific subject matter of the challenged claim, for a court can take account of the inferences and creative steps that a person of ordinary skill in the art would employ. KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 418 (2007) (quoting In re Kahn, 441 F.3d 977, 988 (Fed. Cir. 2006)). Boal teaches using contextual signals such as transaction data, shopping list data, consumer data, offer data, offer tracking data, or product data to develop offer scores 1776. Boal Abstract, ¶¶ 167–169; Fig. 17. Boal teaches the use of a “learning component” to create the offer scores. Boal ¶ 168; Fig. 17, element 1774. Boal additionally teaches the optional use of weighting (contextual weighting component) to create universal scores to further create ranked offer sets. Boal ¶ 169; Fig. 17, element 1781. Here, despite Appellant’s arguments, the Examiner does not explain why one of ordinary skill in the art would interpret the Boal “learning component” as the claimed “logistic regression machine learning method.” Nor does the Examiner explain why one of ordinary skill in the art would Appeal 2018-009059 Application 14/893,606 8 understand Boal’s contextual weighting 1781 to be the claimed “logistic regression machine learning method.” Additionally, the Examiner presents no evidence or articulated reasoning that one of ordinary skill in the art would modify the cited teaching of Boal and employ the claimed “logistic regression machine learning method.” Therefore, we do not sustain the rejection of claim 1, independent claims 6 and 11, which include disputed limitation L2, and dependent claims 2, 5, 7–10, 12, and 14–19. We also do not sustain the rejection of dependent claims 3, 4, 13, and 20, which are rejected over the combination of Lim, Boal, and the additional cited references, which the Examiner does not rely on to cure the deficiency, supra, regarding disputed limitation L2. Because our decision with regard to the disputed limitation is dispositive of the rejections made, we do not address additional arguments raised by Appellant. CONCLUSION The Examiner’s rejection of claims 1, 2, 5–12, and 14–19 under 35 U.S.C. § 103(a) as being unpatentable over Lim and Boal is reversed. The Examiner’s rejection of claims 3 and 4 under 35 U.S.C. § 103 as being unpatentable over Lim, Boal, and Chenthamarakshan is reversed. The Examiner’s rejection of claim 13 under 35 U.S.C. § 103 as being unpatentable over Lim, Boal, and Zeng is reversed. The Examiner’s rejection of claim 13 under 35 U.S.C. § 103 as being unpatentable over Lim, Boal, and Rios is reversed. Appeal 2018-009059 Application 14/893,606 9 DECISION SUMMARY Claims Rejected 35 U.S.C. § Reference(s)/Basis Affirmed Reversed 1, 2, 5–12, 14–19 103(a) Lim, Boal 1, 2, 5–12, 14–19 3, 4 103(a) Lim, Boal, Chenthamarakshan 3, 4 13 103(a) Lim, Boal, Zeng 13 20 103(a) Lim, Boal, Rios 20 Overall Outcome: 1–20 REVERSED Copy with citationCopy as parenthetical citation