Facebook, Inc.Download PDFPatent Trials and Appeals BoardMay 21, 20212020000527 (P.T.A.B. May. 21, 2021) 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. 14/616,543 02/06/2015 Rituraj Kirti 2006.032US1 1171 160546 7590 05/21/2021 Mannava & Kang, P. C. 3201 Jermantown Road Suite 525 Fairfax, VA 22030 EXAMINER STROUD, CHRISTOPHER ART UNIT PAPER NUMBER 3688 NOTIFICATION DATE DELIVERY MODE 05/21/2021 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): docketing@mannavakang.com fb-pdoc@fb.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE ____________________ BEFORE THE PATENT TRIAL AND APPEAL BOARD ____________________ Ex parte RITURAJ KIRTI, SUE ANN HONG, and LEON R. CHO ____________________ Appeal 2020-000527 Application 14/616,543 Technology Center 3600 ____________________ Before JOSEPH L. DIXON, DAVID M. KOHUT, and JON M. JURGOVAN, Administrative Patent Judges. JURGOVAN, Administrative Patent Judge. DECISION ON APPEAL Appellant1 seeks review under 35 U.S.C. § 134(a) from a final rejection of claims 1, 3–10, 12–17, 19, and 20.2 We have jurisdiction under 35 U.S.C. § 6(b). We AFFIRM.3 1 We use the word “Appellant” to refer to “applicant” as defined in 37 C.F.R. § 1.42. The real party in interest is Facebook, Inc. (Appeal Br. 2.) 2 Claims 2, 11, and 18 were canceled. (Final Act. 1.) 3 Our Decision refers to the Specification (“Spec.”) filed February 6, 2015, Final Office Action (“Final Act.”) mailed March 14, 2019, the Appeal Brief (“Appeal Br.”) filed July 25, 2019, the Examiner’s Answer (“Ans.”) mailed August 29, 2019, and the Reply Brief (“Reply Br.”) filed October 29, 2019. Appeal 2020-000527 Application 14/616,543 2 CLAIMED INVENTION The claims are directed to computer-implemented methods and program products for “determining a number of cluster groups associated with content identifying users eligible to receive the content” (e.g., advertisement content including text, images, videos, and/or animation), and increasing a number of users eligible to be presented with the content. (Spec. ¶ 26; Title (capitalization altered); Abstr.) Independent claim 1, reproduced below, is illustrative of the claimed subject matter: 1. A computer-implemented method comprising: receiving an advertisement request including a first set of targeting criteria and a second set of targeting criteria at an online system, wherein the advertisement request further includes first advertisement content associated with the first set of targeting criteria and second advertisement content associated with the second set of targeting criteria; retrieving characteristics of a plurality of users of the online system; identifying a first targeting group that comprises a first subset of users from the plurality of users of the online system having one or more characteristics that satisfy the first set of targeting criteria; identifying a second targeting group that comprises a second subset of users from the plurality of users of the online system having one or more characteristics that satisfy the second set of targeting criteria; generating a first cluster group by applying a first cluster model to the characteristics of at least some of the plurality of users of the online system who are not in the first targeting group, where the first cluster model is a machine learning model that is trained to determine membership in the first cluster group using the first subset of users as a training set, wherein the first cluster model outputs a score for inclusion in the first cluster group based on the users in the first targeting group, and wherein at Appeal 2020-000527 Application 14/616,543 3 least one of the at least some of the plurality of users of the online system who are not in the first targeting group is added to the first cluster group responsive to the score for inclusion output by application of the first cluster model to characteristics of the at least one user exceeding a threshold score; generating a second cluster group by applying a second cluster model to the characteristics of at least some of the plurality of users of the online system who are not in the second targeting group, where the second cluster model is a machine learning model that is trained to determine membership in the second cluster group using the second subset of users as a training set, wherein the second cluster model outputs a score for inclusion in the second cluster group based on the users in the second targeting group wherein at least one of the at least some of the plurality of users of the online system who are not in the second targeting group is added to the second cluster group responsive to the score for inclusion output by application of the second cluster model to characteristics of the at least one user exceeding a threshold score; determining a first amount of overlap between the first cluster group and the second cluster group, the first amount of overlap based at least in part on a number of users included in both the first cluster group and the second cluster group and comprising at least one user included in both the first cluster group and the second cluster group; determining whether the first amount of overlap equals or exceeds a first threshold amount of overlap; responsive to the first amount equaling or exceeding the first threshold amount of overlap: combining the first and second cluster groups into an overall cluster group; and labeling each user in the overall cluster group with a classifier, wherein users in the overall cluster group from the first cluster group are each labeled with a first classifier and users in the overall cluster group from the second cluster group are each labeled with a second classifier; and sending for presentation to at least one user in the overall cluster group either the first advertisement content or the second advertisement content based on the classifier labeling the at least Appeal 2020-000527 Application 14/616,543 4 one user, wherein users labeled with the first classifier are sent the first advertisement content and users labeled with the second classifier are sent the second advertisement content. (Appeal Br. 21–23 (Claims App.).) REJECTIONS4 & REFERENCES Claims 1, 3–10, 12–17, 19, and 20 stand rejected under 35 U.S.C. § 112(a) for failing to comply with the written description requirement. (Final Act. 2–3.) Claims 1, 3–10, 12–17, 19, and 20 stand rejected under 35 U.S.C. § 101 as directed to non-statutory subject matter. (Final Act. 4– 7.) Claims 1, 8–10, and 17 stand rejected under 35 U.S.C. § 103 based on Zohar et al. (US 2014/0304066 A1, published Oct. 9, 2014) (“Zohar”), Chang et al. (US 2011/0106611 A1, published May 5, 2011) (“Chang”), Joa et al. (US 2014/0257933 A1, published Sept. 11, 2014) (“Joa”), and Li et al. (US 2013/0124298 A1, published May 16, 2013) (“Li”). (Final Act. 8–12.) Claims 3–7, 12–16, 19, and 20 stand rejected under 35 U.S.C. § 103 based on Zohar, Chang, Joa, Li, and Gunawardana et al. (US 2009/0327032 A1, published Dec. 31, 2009) (“Gunawardana”). (Final Act. 13–15.) 4 Claims 17, 19, and 20 were rejected under 35 U.S.C. § 112(b), as being indefinite. (Final Act. 4.) However, this rejection was withdrawn in the Examiner’s Advisory Action (mailed May 15, 2019), and is no longer pending on appeal. (Advisory Act. 1.) Appeal 2020-000527 Application 14/616,543 5 ANALYSIS Standard of Review We undertake a limited de novo review of the appealed rejections for error based upon the issues identified by Appellant, and in light of the arguments and evidence produced thereon. Ex parte Frye, 94 USPQ2d 1072 (BPAI 2010) (precedential). Section 112(a) Rejection Section 112(a) requires that [t]he specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains . . . to make and use the same.5 The written description must reasonably convey to those skilled in the art that the inventor had possession of the claimed subject matter as of the filing date. Ariad Pharm., Inc. v. Eli Lilly & Co., 598 F.3d 1336, 1351 (Fed. Cir. 2010) (en banc); D Three Enterprises, LLC v. SunModo Corp., 890 F.3d 1042, 1047 (Fed. Cir. 2018). The Examiner finds the Specification lacks written description support for “a machine learning model that is trained to determine membership in the first [or second] cluster group using the first [or second] subset of users as a training set,” as recited in claim 1, and similarly in claims 10 and 17. 5 The Leahy-Smith America Invents Act, Pub. L. No. 112-29, 125 Stat. 284 (September 16, 2011) (“AIA”), included revisions to 35 U.S.C. § 112 that became effective on September 16, 2012. Because the present patent application was filed on February 6, 2015, AIA § 112 applies to this application. Appeal 2020-000527 Application 14/616,543 6 (Final Act. 3; Ans. 3–5.) In particular, the Examiner asserts that Appellant’s Specification “has not supplied the machine learning algorithm nor does it state how the machine learning algorithm determines a score for determining membership in a cluster” (Final Act. 3) and merely provides a generic statement that basically says we can do cluster modeling with machine learning. . . . mak[ing] reference to required weighting, input, hidden, and output layers without providing any information what so ever as to how these functions would be achieved. One of ordinary skill in the art would simply not be able to replicate the specific algorithm required to implement the invention. While the concept of machine learning may be well known to one of ordinary skill in the art, the specifically claimed algorithm has not been supplied and thus it has not been shown that applicant has possession of the invention (Ans. 4). Appellant argues the Specification provides written description support under § 112 for the claimed “machine learning model that is trained to determine membership in the first cluster group using the first subset of users as a training set” and “machine learning model that is trained to determine membership in the second cluster group using the second subset of users as a training set.” (Appeal Br. 10–11 (citing Spec. ¶¶ 6, 53–54, 62– 65); Reply Br. 2–4.) In particular, Appellant argues the Specification demonstrates Appellant possessed the claimed machine learning model(s) because the Specification provides sufficient implementation details— including the model’s training input, the desired model output (a prediction of whether a user is in a cluster group), and how the model is used to cluster users—for one of skill in the art . . . to build the claimed cluster models. One of skill in the art can select any appropriate type of model (e.g., Appeal 2020-000527 Application 14/616,543 7 a neural network or a linear regression). The algorithms for training known machine learning models (e.g., backpropagating a neural network or fitting a linear regression) are well known. (Appeal Br. 10–11; Reply Br. 4–5.) We agree with Appellant’s arguments. As Appellant explains, machine learning algorithms are known in the art, such algorithms known to input and fit training data and determine scores for new data based on the training. (Appeal Br. 9–10; Reply Br. 4.) Claim 1 broadly recites a machine learning model that is trained to determine membership using a set of users as a training set, and Appellant’s Specification sufficiently discloses a machine learning model (see Spec. ¶¶ 70, 73, 74, “unsupervised machine learning algorithm,” “artificial neural network”) that is trained (using, e.g., “weights of connections between input, hidden, and output layers of the neural network,” “parameters . . . determined using an iterative process to minimize a cost function over the training data,” and “weights applied to various characteristics of a user,” see Spec. ¶¶ 62, 70, 73) using training set(s) of users (defined by personal characteristics/interests, see Spec. ¶¶ 20, 22, 53, 70, 72–74), to determine membership (e.g., users’ affinity or interest for various content types, see Spec. ¶¶ 6, 53, 62–63, 69, 74). (Appeal Br. 9–10; Reply Br. 3–4; see Fonar Corp. v. Gen. Elec. Co., 107 F.3d 1543, 1549 (Fed. Cir. 1997) (“As a general rule . . . writing code for [] software is within the skill of the art, not requiring undue experimentation, once its functions have been disclosed. . . . [and] flow charts or source code listings are not a requirement for adequately disclosing the functions of software.”); Hybritech Inc. v. Monoclonal Antibodies, Inc., 802 F.2d 1367, 1379–80, 1384 (Fed. Cir. 1986) Appeal 2020-000527 Application 14/616,543 8 (information that is conventional or well known in the art need not be described in detail in the specification).) We, therefore, find there is written description support for the claimed machine learning model(s) trained to determine membership in clusters using user groups as training sets, and do not sustain the Examiner’s § 112(a) rejection of independent claims 1, 10, and 17, and claims 3–9, 12– 16, 19, and 20, depending therefrom. § 101 Rejection Patent eligibility is a question of law that is reviewable de novo. Dealertrack, Inc. v. Huber, 674 F.3d 1315, 1333 (Fed. Cir. 2012). Accordingly, we review the Examiner’s § 101 determinations concerning patent eligibility under this standard. Patentable subject matter is defined by 35 U.S.C. § 101, as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title. In interpreting this statute, the Supreme Court emphasizes that patent protection should not preempt “the basic tools of scientific and technological work.” Gottschalk v. Benson, 409 U.S. 63, 67 (1972) (“Benson”); Mayo Collaborative Servs. v. Prometheus Labs., Inc., 566 U.S. 66, 71 (2012) (“Mayo”); Alice Corp. v. CLS Bank Int’l, 573 U.S. 208, 217–18 (2014) (“Alice”). The rationale is that patents directed to basic building blocks of technology would not “promote the progress of science and useful arts” under the U.S. Constitution, Article I, Section 8, Clause 8, but instead would impede it. Accordingly, laws of nature, natural phenomena, and abstract Appeal 2020-000527 Application 14/616,543 9 ideas, are not patent-eligible subject matter. Thales Visionix Inc. v. United States, 850 F.3d 1343, 1346 (Fed. Cir. 2017) (citing Alice, 573 U.S. at 216– 17). The Supreme Court set forth a two-part test for subject matter eligibility in Alice (573 U.S. at 217–19). The first step is to determine whether the claim is directed to a patent-ineligible concept. Id. (citing Mayo, 566 U.S. at 76–77). If so, then the eligibility analysis proceeds to the second step of the Alice/Mayo test in which we “examine the elements of the claim to determine whether it contains an ‘inventive concept’ sufficient to ‘transform’ the claimed abstract idea into a patent-eligible application.” Alice, 573 U.S. at 221 (quoting Mayo, 566 U.S. at 72, 79). There is no need to proceed to the second step, however, if the first step of the Alice/Mayo test yields a determination that the claim is directed to patent eligible subject matter. The Patent Office has revised its guidance for how to apply the Alice/Mayo test in the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (January 7, 2019) (the “Revised Guidance”). Under the Revised Guidance, we first look to whether the claim recites: (1) any judicial exceptions, including certain groupings of abstract ideas (i.e., mathematical concepts, mental processes, or certain methods of organizing human activity such as a fundamental economic practice or managing personal behavior or relationships or interactions between people); and (2) additional elements that integrate the judicial exception into a practical application (see Manual of Patent Examining Procedure (“MPEP”) § 2106.05(a)–(c), (e)–(h)). 84 Fed. Reg. at 51–52, 55. A claim that integrates a judicial exception into a practical application applies, relies on, or uses the judicial exception in a manner that imposes a Appeal 2020-000527 Application 14/616,543 10 meaningful limit on the judicial exception, such that the claim is more than a drafting effort designed to monopolize the judicial exception. Revised Guidance, 84 Fed. Reg. at 54. When the judicial exception is so integrated, then the claim is not directed to a judicial exception and is patent-eligible under § 101. Revised Guidance, 84 Fed. Reg. at 54. Only if a claim (1) recites a judicial exception and (2) does not integrate that exception into a practical application, do we then evaluate whether the claim provides an inventive concept. Revised Guidance, 84 Fed. Reg. at 56; Alice, 573 U.S. at 217–19, 221. Evaluation of the inventive concept involves consideration of whether an additional element or combination of elements (1) adds a specific limitation or combination of limitations that are not well-understood, routine, conventional activity in the field, which is indicative that an inventive concept may be present; or (2) simply appends well-understood, routine, conventional activities previously known to the industry, specified at a high level of generality, to the judicial exception, which is indicative that an inventive concept may not be present. USPTO Step 1 – Do the claims recite a statutory category of invention? Claims 1, 3–10, and 12–16 recite computer-implemented methods that qualify as “processes” under 35 U.S.C. § 101, so USPTO Step 1 of the patent eligibility analysis is satisfied. Claims 17, 19, and 20 recite computer program products which qualify as articles of manufacture under 35 U.S.C. § 101. Accordingly, the claims satisfy USPTO Step 1. We proceed in our analysis to the first step of Alice/Mayo. Appeal 2020-000527 Application 14/616,543 11 Alice/Mayo—Step 1 (Abstract Idea) USPTO Step 2A–Prongs 1 and 2 identified in the Revised Guidance USPTO Step 2A—Prong 1 (Does the Claim Recite a Judicial Exception?) Turning to the first step of the Alice inquiry (Step 2A, Prong 1 of the Revised Guidance), the Examiner finds independent claim 1, and similarly, independent claims 10 and 17, are directed to an abstract idea because the claims recite a process that would be performed precisely the same way with or without a computer. The claims involve the steps of receiving ad requests with targeting criteria, retrieving characteristics of users, identifying a targeting group with those characteristics, analyzing other users to see if they are similar enough to be put in a cluster, examining clusters to see if there is a significant amount of overlap in the clusters, combine the groups if enough overlap exists, labeling the users of the cluster, and then providing an advertisement to a member of the cluster based on the identifier. All of these steps do not require the use of a computer. They would be performed exactly the same with in the human mind or with pencil and paper. Thus all of the grouping and data analysis are in fact a mental process. (Ans. 6.) The Examiner reasons that [the claimed] process . . . under its broadest reasonable interpretation, cover[sic] performance of the limitations in the mind but for the recitation of a generic computing environment of “an online system” and generic mention of “machine learning.” Machine learning is in and of itself simply an iterative process of reevaluating as more data comes in. In particular, recited this generically it is little more than automating the same process that can be done by a human in the exact same manner. But for the recitation of the “online system” the steps can be Appeal 2020-000527 Application 14/616,543 12 performed mentally in the human mind or by a human analog with pen and paper. (Final Act. 6–7.) Appellant contends the Examiner erred in rejecting claims 1, 10, and 17 under 35 U.S.C. § 101 as directed to non-statutory subject matter because the claims are not directed to an abstract idea. (Appeal Br. 11–16 (referencing “Claims 1, 10, and 17” and submitting arguments for limitations recited in claim 1); Reply Br. 5–7.) Particularly, Appellant contends “[t]he claims . . . do not recite a mental process,” and “integrate the machine learning technique into a practical application” for “addressing overlap among cluster groups, which is a problem that arises due to the use of machine learning to generate the expanded cluster.” (Appeal Br. 12, 15– 16; see also Reply Br. 5–6.) Appellant further argues “the claimed invention is not merely ‘machine learning’ in the abstract, but rather an integration of the concept into a practical application” that “uses machine learning to find an expanded cluster group of users that are similar to an existing seed group of users (a targeting group) . . . for multiple seed groups.” (Reply Br. 6–7.) Appellant’s arguments do not persuade us that the Examiner erred in finding claim 1 (and claims 10 and 17 reciting similar limitations) recites an abstract idea and, therefore, we concur with the Examiner’s conclusion that the claim recites an abstract idea of distributing targeted advertisements by a process performable in the human mind or with pen and paper. (Final Act. 4–7; Ans. 6–8.) Under its broadest reasonable interpretation, claim 1 recites an abstract mental process of determining user groups to which ads are to be distributed. (Final Act. 6; Ans. 6–7.) In particular, claim 1 recites an abstract mental process of grouping users to receive targeted ads by: Appeal 2020-000527 Application 14/616,543 13 collecting/gathering information (claimed “receiving an advertisement request including a first set of targeting criteria and a second set of targeting criteria” as well as “first advertisement content associated with the first set of targeting criteria and second advertisement content associated with the second set of targeting criteria,” and “retrieving characteristics of a plurality of users”), analyzing the information (claimed “identifying a first targeting group that comprises a first subset of users from the plurality of users . . . having one or more characteristics that satisfy the first set of targeting criteria,” “identifying a second targeting group that comprises a second subset of users from the plurality of users . . . having one or more characteristics that satisfy the second set of targeting criteria,” “generating a first [or second] cluster group by applying a first [or second] cluster model to the characteristics of at least some of the plurality of users . . . who are not in the first [or second] targeting group” with the first/second cluster model outputting “a score for inclusion in the first [or second] cluster group based on the users in the first [or second] targeting group,” the first and second cluster models being a model “trained to determine membership in the first [or second] cluster group using the first [or second] subset of users as a training set,” adding “at least one of the at least some of the plurality of users . . . who are not in the first [or second] targeting group . . . to the first [or second] cluster group” responsive to a score “output by application of the first [or second] cluster model to characteristics of the at least one user exceeding a threshold score,” “determining a first amount of overlap between the first cluster group and the second cluster group,” “determining whether the first amount of overlap equals or exceeds a first threshold amount of overlap,” and “responsive to the first amount equaling or Appeal 2020-000527 Application 14/616,543 14 exceeding the first threshold amount of overlap: combining the first and second cluster groups into an overall cluster group” and “labeling each user in the overall cluster group with a classifier, wherein users in the overall cluster group from the first cluster group are each labeled with a first classifier and users in the overall cluster group from the second cluster group are each labeled with a second classifier”), and providing results of the collection and analysis (“sending for presentation to at least one user in the overall cluster group either the first advertisement content or the second advertisement content based on the classifier labeling” whereby “users labeled with the first classifier are sent the first advertisement content and users labeled with the second classifier are sent the second advertisement content”). That is, although claim 1 recites users of “an online system,” advertisement requests received at “the online system,” and “a machine learning model” for the first and second cluster models, the underlying operations recited in the claim are acts that could be performed mentally and by pen and paper, without the use of an online system or machine learning. (Ans. 6.) For example, a person could (i) manually retrieve and analyze sets of targeting criteria and characteristics of users, (ii) manually identify first and second targeting groups, use cluster model(s) trained to determine membership in cluster groups to generate first and second cluster groups by applying the models to characteristics of some users that are not in particular targeting group(s), and manually determine scores for inclusion in respective cluster groups, (iii) manually add at least one user not in the first or second targeting group to the first or second cluster group, based on the scores, (iv) manually determine an overlap between cluster groups and, if the overlap Appeal 2020-000527 Application 14/616,543 15 equals or exceeds a threshold, manually generate an overall cluster group by combining cluster groups, and manually label users in the overall cluster group with classifiers based on the originating cluster group, and (v) manually assign first or second ads to at least one user in the overall cluster group, based on the user’s classifier. (See Ans. 6, 8; Final Act. 6.) Our reviewing court has concluded that mental processes include similar concepts of collecting, manipulating, and providing, data. See Intellectual Ventures I LLC v. Capital One Fin. Corp., 850 F.3d 1332, 1340 (Fed. Cir. 2017) (the Federal Circuit held “the concept of . . . collecting data, . . . recognizing certain data within the collected data set, and . . . storing that recognized data in a memory” ineligible); and Electric Power Grp., LLC v. Alstom S.A., 830 F.3d 1350, 1353 (Fed. Cir. 2016) (merely selecting information, by content or source, for collection, analysis, and display does nothing significant to differentiate a process from ordinary mental processes). Claim 1’s “online system” merely automates data input and output, and claim 1’s recitation of a “machine learning model” merely references an automation of actions that are manually performable (e.g., actions of determining membership in clusters based on characteristics of pre-selected training sets). However, mental processes remain unpatentable even when automated to reduce the burden on the user of what once could have been done with pen and paper. See CyberSource Corp. v. Retail Decisions, Inc., 654 F.3d 1366, 1375 (Fed. Cir. 2011) (“That purely mental processes can be unpatentable, even when performed by a computer, was precisely the holding of the Supreme Court in Gottschalk v. Benson.”). Thus, Appellant’s arguments that claims 1, 10, and 17 recite elements that cannot be characterized as mental steps or performed mentally, and that Appeal 2020-000527 Application 14/616,543 16 the claims recite more than an abstract idea, have not persuaded us the Examiner erred in finding the claims also recite an abstract idea.6 We now turn to USPTO Step 2A, Prong 2, of the Revised Guidance to determine whether the abstract idea is integrated into a practical application. See Revised Guidance, 84 Fed. Reg. at 54–55. USPTO Step 2A—Prong 2 (Integration into Practical Application) Under Revised Step 2A, Prong Two of the Revised Guidance, we discern no additional element (or combination of elements) recited in Appellant’s claims 1, 10, and 17 that may have integrated the judicial exception into a practical application. See Revised Guidance, 84 Fed. Reg. at 54–55. We recognize that claim 1 (and similarly, claims 10 and 17) includes additional elements such as an operational connection to an “online system,” and the use of cluster models that are “a machine learning model that is trained to determine membership in the . . . cluster group using the first [or second] subset of users as a training set.” Furthermore, our review of Appellant’s Specification finds that the terms “online system” and “machine learning model” are nominal. Appellant’s Specification indicates that the “online system” and “machine learning model” (see Spec. ¶¶ 4, 19, 28, 44, 70, 73–74) of claim 1 do not recite specific types of additional elements or their operations, thereby requiring no improvements to computing or data processing technologies. For example, Appellant’s claimed additional elements (“online system,” and cluster models that are 6 As the Examiner additionally notes, claim 1 also recites an abstract idea within a grouping of “Certain methods of organizing human activity” described in the Revised Guidance. (See Revised Guidance, 84 Fed. Reg. at 52; Ans. 6, Final Act. 6–7.) Appeal 2020-000527 Application 14/616,543 17 “machine learning model[s]”) do not: (1) improve the functioning of a computer or other technology; (2) are not applied with any particular machine (except for generic computing elements); (3) do not effect a transformation of a particular article to a different state; and (4) are not applied in any meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than a drafting effort designed to monopolize the exception. (See MPEP §§ 2106.05(a)–(c), (e)–(h); Ans. 7–8; Final Act. 7.) As a result, the claim’s recited additional elements are not indicative of “integration into a practical application,” and are not enough to distinguish the steps of claim 1 from describing a mental process. See Revised Guidance, 84 Fed. Reg. at 54–55. Appellant argues claim 1 is not directed to a mental process because the additional [claim] elements of “generating a first cluster group . . . ,” “generating a second cluster group . . . ,” “determining a first amount of overlap . . . ,” “determining whether the first amount of overlap equals . . . ,” “combining the first and second cluster groups . . . ,” and “labeling each user . . .” integrate the machine learning technique into a practical application. (Appeal Br. 15.) Appellant asserts the “practical application [is] for addressing overlap among cluster groups, which is a problem that arises due to the use of machine learning to generate the expanded cluster.” (Id.) Appellant further argues [n]o one with any understanding of neural networks would suggest that a neural network of any kind could be practically performed in the human mind. . . . Similarly, no one with any understanding of machine learning in general would suggest that the steps of the present claims are practically performable in the Appeal 2020-000527 Application 14/616,543 18 human mind. (Reply Br. 6.) Appellant’s arguments are unpersuasive for the following reasons. The claimed “generating a first cluster group,” “generating a second cluster group,” “determining a first amount of overlap,” “determining whether the first amount of overlap equals or exceeds a first threshold,” “combining the first and second cluster groups,” and “labeling each user in the overall cluster group” are operations readily performable in the human mind or with pen and paper, and are part of the abstract idea (a mental process, as discussed supra at USPTO Step 2A, Prong 1). (See Ans. 6, 8; Final Act. 6– 7.) With respect to the claimed limitations requiring the first and second cluster models to be “a machine learning model that is trained to determine membership in the first [or second] cluster group using the first [or second] subset of users as a training set,” we agree with the Examiner that these limitations are insufficient to integrate the abstract idea into a practical application. (See Ans. 7–8; Final Act. 7.) As explained by the Examiner, the machine learning model limitations in Appellant’s claims are not novel, and can be reasonably seen as the conventional application of well-known machine learning concepts to build and train a model using a training set. (Ans. 7–8; Final Act. 7.) We agree with the Examiner that claim 1 does not recite a specific implementation of machine learning that “ties it to a technological process”; rather, the claim recites the generic use of machine learning to train a model to determine group membership based on training sets, and [the] generic recitation of “machine learning” amounts to no more than an iterative process in which a model is updated as Appeal 2020-000527 Application 14/616,543 19 new information is obtained and analyzed. . . . under broadest reasonable interpretation a generic use of machine learning implies nothing more than updating a model as more data becomes available. . . . Further a generic recitation of “machine learning” is equivalent to the words “apply it” on a computer or simply instruct the computer to implement the abstract idea. (Ans. 6–7.) In addition, Appellant’s Specification describes the use of machine learning in general terms, without explaining how Appellant’s “machine learning model” would be different from known machine learning and artificial neural network algorithms. (See, e.g., Spec. ¶¶ 70, 73–74.) Because Appellant’s Specification provides a high-level and general description of machine learning requiring no specialized implementation, we agree with the Examiner that the “machine learning” limitations in claims 1, 10, and 17 do not evidence any improvements to the functioning of a computer or any other technology or technical field. (Ans. 6–8; see MPEP § 2106.05(a).) As noted above, Appellant also argues the claims “integrate the machine learning technique into a practical application” for “addressing overlap among cluster groups, which is a problem that arises due to the use of machine learning to generate the expanded cluster.” (See Appeal Br. 15.) As noted by the Examiner, however, “the problem addressed has nothing [whatsoever] to do with machine learning or computer related issues. The problem alleged by appellant is not a machine learning issue, it is a grouping issue which is related to the abstract idea.” (Ans. 7.) Appellant’s Specification does not explain why the problem of cluster group overlap would be an exclusive result of using machine learning, or “how [this] Appeal 2020-000527 Application 14/616,543 20 problem is unique or a result of machine learning.” (Ans. 7–8.) In addition, Appellant’s asserted improvement (addressing overlap among cluster groups subjected to targeted advertising) is not an improvement to machine learning or other technology, rather, it is an improvement to targeted advertising and the use of advertising resources. (See Spec. ¶¶ 54–55.) Appellant also seeks to analogize claim 1 to a hypothetical claim identified as eligible in the Office’s Patent Eligibility Example 39 (Method for Training a Neural Network for Facial Detection).7 (Appeal Br. 12–13; Reply Br. 5–6.) As the Examiner points out, however, “[t]here are several major differences between example 39 and the current claims.” (Ans. 5.) In particular, example 39 is directed to a process of digital facial detection from images. This is a process rooted in computers similar to the findings in McRo. . . . the claims in example 39 could not be considered a mental process as the process was for digital facial recognition and could not be considered organizing human activities because it would not be performed in the same manner if a human being were to attempt the task without a computer. In stark contrast, the current claims provide a process that would be performed precisely the same way with or without a computer. . . . With regard to the added buzzword of “machine learning,” appellant has not provided any meaningful limitation that adds any significance to its use that ties it to a technological process. . . . A generic recitation of “machine learning” amounts to no more than an iterative process in which a model is updated as new information is obtained and analyzed. (Ans. 5–6 (emphasis added).) We agree with the Examiner’s findings. 7 Located at: https://www.uspto.gov/sites/default/files/documents/101_examples_37to42_ 20190107.pdf Appeal 2020-000527 Application 14/616,543 21 Appellant further argues that claim 1 recites additional elements that integrate a judicial exception into a practical application because the claim is similar to a hypothetical claim 1 identified as eligible in the Office’s Patent Eligibility Example 42 (Method for Transmission of Notifications When Medical Records Are Updated).8 (Appeal Br. 15–16; Reply Br. 6.) That hypothetical claim, however, recited a method comprising steps for converting non-standardized information dependent on the hardware and software platform used by a user providing the information, into a standardized format, storing the standardized updated information about a patient’s condition in a collection of medical records in the standardized format, and automatically generating a message containing the updated information about the patient’s condition by a content server whenever updated information has been stored. Thus, unlike Appellant’s claim 1 (which recites a generic use of machine learning for clustering data, e.g., user descriptors), the hypothetical method recites a combination of additional elements integrating a practical application—one that provides a specific improvement over prior art medical records management systems, by allowing remote users to share information in real time in a standardized format regardless of the format in which the information was input by users. As noted by the Examiner, “[u]nlike example 42 the claimed invention provides no improvements over prior systems and the problem addressed has nothing what so ever to do with machine learning or computer related issues” as the “problem alleged by appellant is not a machine learning issue, 8 Located at: https://www.uspto.gov/sites/default/files/documents/101_examples_37to42_ 20190107.pdf Appeal 2020-000527 Application 14/616,543 22 it is a grouping issue which is related to the abstract idea.” (Ans. 7–8.) We agree with the Examiner. Accordingly, under Step 2A, Prong 2, we conclude that claim 1, and similarly worded independent claims 10 and 17 argued for the same reasons do not recite “additional elements that integrate the [judicial] exception into a practical application,” and are directed to an abstract idea in the form of a mental process. Revised Guidance, 84 Fed. Reg. at 52, 54. Therefore, we proceed to USPTO Step 2B, The Inventive Concept. Alice/Mayo—Step 2 (Inventive Concept) USPTO Step 2B identified in the Revised Guidance USPTO Step 2B of the Alice two-step framework requires us to determine whether any element, or combination of elements, in the claim is sufficient to ensure that the claim amounts to significantly more than the judicial exception. Alice, 573 U.S. at 221; see also Revised Guidance, 84 Fed. Reg. at 56. As discussed above, claim 1 includes additional elements such as an “online system” and cluster models that are “machine learning model[s].” We agree with the Examiner’s findings that the additional elements of claim 1 (and similar ones in claims 10 and 17), when considered individually and in an ordered combination, correspond to nothing more than generic and well-known components and operations used to implement the abstract idea. (See Final Act. 7; Ans. 7–8.) In other words, we find that the additional elements, as claimed, are well-understood, routine, and conventional and “behave exactly as expected according to their ordinary use.” (See In re TLI Commc’ns LLC Patent Litig., 823 F.3d 607, 615 (Fed. Cir. 2016); Final Act. 7; Ans. 7–8.) Thus, implementing the abstract idea Appeal 2020-000527 Application 14/616,543 23 with these generic and well-known elements “fail[s] to transform that abstract idea into a patent-eligible invention.” Alice, 573 U.S. at 221. Therefore, we agree with the Examiner that claims 1, 10, and 17 do not provide significantly more than the abstract idea itself. Therefore, because claims 1, 10, and 17 are directed to the abstract idea of a mental process, and do not provide significantly more than the abstract idea itself, we agree with the Examiner that claims 1, 10, and 17 are ineligible for patenting. We, therefore, sustain the Examiner’s § 101 rejection of independent claims 1, 10, and 17, and dependent claims 3–9, 12–16, 19, and 20 not separately argued. See 37 C.F.R. § 41.37(c)(1)(iv) (“the failure of appellant to separately argue claims which appellant has grouped together shall constitute a waiver of any argument that the Board must consider the patentability of any grouped claim separately.”). § 103 Rejection Title 35, section 103, provides: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. The question of obviousness is resolved on the basis of underlying factual determinations including: (1) the scope and content of the prior art; (2) any differences between the claimed subject matter and the prior art; (3) the level of ordinary skill in the art; and (4) where present, objective evidence of nonobviousness. Graham v. John Deere Co., 383 U.S. 1, 17–18 (1966). Appeal 2020-000527 Application 14/616,543 24 With respect to independent claim 1, the Examiner finds Zohar’s segments (including behavioral segments, segment combinations, and unified combinations of segments) teach the claimed “first targeting group” and “second targeting group.” (Final Act. 9 (citing Zohar ¶¶ 16, 84–87, 92– 93, 105).) The Examiner also finds Zohar teaches the claimed “determining a first amount of overlap between the first cluster group and the second cluster group . . . based at least in part on a number of users included in both the first cluster group and the second cluster group,” because Zohar determines a percentage overlap between users included in a behavioral segment and users included in a unified combination. (Final Act. 9–10 (citing Zohar ¶¶ 114–117, 181–183) (emphasis omitted)). The Examiner acknowledges “Zohar does not expressly teach generating cluster groups including users not in targeting segments” but finds Chang’s expansion of targeted segments by complementary user segments teaches the claimed generation of first and second cluster groups by applying cluster models to characteristics of at least some users who are not in first or second targeting groups. (Final Act. 10 (citing Chang ¶¶ 14, 26–28); Ans. 8–9.) The Examiner asserts “[i]f you combine the groups from Zohar with the expansion of groups from Chang you have the cluster groups of the claims,” as [i]t would have been obvious to one of ordinary skill in the art at the time of filing of the invention to include other members in a group that do not meet the targeting criteria as taught by Chang in order to expand the reach of targeted advertisements to other potential interested groups. . . . Further, since each element of the combination performs the same functions individually as it does in combination one of ordinary skill in the art would find the combination to yield predictable results. Appeal 2020-000527 Application 14/616,543 25 (Ans. 9; Final Act. 11.) We do not agree. We agree with Appellant that Zohar and Chang, alone or in combination, fail to teach or suggest determining an overlap between first and second cluster groups obtained by expanding membership to users that are similar to, but outside of, user sets of first and second targeting groups, as recited in claim 1. (Appeal Br. 17–19; Reply Br. 8–9.) Although the Examiner has identified a teaching of generating cluster groups by expanding initial target groups (in Chang reference), and a teaching of determining an overlap between targeting groups (in Zohar), the Examiner has not provided a reason to combine the teachings of Zohar and Chang to (1) first generate cluster groups by expanding initial targeting groups, and then (2) determine an overlap between such cluster groups obtained by expanding targeting groups. (Appeal Br. 18–19; Reply Br. 8–9.) Initially, we note that—contrary to the Examiner’s assertion that Zohar teaches the claimed “determining a first amount of overlap between the first cluster group and the second cluster group” (see Final Act. 9)—Zohar does not determine an overlap between cluster groups like those recited in claim 1. (Appeal Br. 18–19; Reply Br. 8–9.) Rather, Zohar determines an overlap between a “behavioral segment” and a “unified combination,” neither the “behavioral segment” nor the “unified combination” being cluster groups obtained by expanding membership to users that are similar to, but outside of, user sets of first and second targeting groups (as required by claim 1). (See Zohar ¶¶ 117, 181–183; Appeal Br. 18–19; Reply Br. 8–9.) Zohar’s behavioral segment is a simple targeting group of users (e.g., users having a particular interest), and Zohar’s unified combination is a combination of segments (e.g., groups of users Appeal 2020-000527 Application 14/616,543 26 having the same gender, age range, interests, etc.). (See Zohar ¶¶ 4–5, 21 (“combine two or more qualified [segment] combinations to create, for example, a ‘unified combination’”), 84–88 (“[a] jazz interest segment may be an example of a ‘behavioral’ segment and can be combined with other types of segments, e.g. user location, or user age segments to form segment combinations”), 90–91 (describing “segments definitions, such as behavioral segments, e.g. ‘likes to swim’, or ‘surfs the Web in late evening’”).) Thus, Zohar does not determine an overlap as claimed (i.e., between cluster groups obtained by expanding a membership to users that are similar to, but outside of, user sets of identified targeting groups), and neither does Chang. (Appeal Br. 18–19; Reply Br. 8–9.) In addition, a skilled artisan would not apply Zohar’s overlap calculation (between a behavioral segment and a unified combination) to expanded cluster groups (e.g., of users not in Chang’s initial targeted segments), because the goal of Zohar’s overlap calculation is to check whether a behavioral segment’s definition is outdated or non-relevant (e.g., does not create a user group with predictable responses to targeted advertising) and should be deleted. (See Zohar ¶¶ 7, 113, 116–117, 168, 181–185.) The Examiner’s findings that “[i]f you combine the groups from Zohar with the expansion of groups from Chang you have the cluster groups of the claims” and “one of ordinary skill in the art would find the combination to yield predictable results” (see Ans. 9 and Final Act. 11) do not adequately explain why the references would have been combined as in the disputed limitation of claim 1 (i.e., to determine overlap between cluster groups obtained by expanding membership to users that are similar to, but outside of, user sets of previously identified targeting groups). See Belden Inc. v. Berk–Tek LLC, Appeal 2020-000527 Application 14/616,543 27 805 F.3d 1064, 1073 (Fed. Cir. 2015) (“[O]bviousness concerns whether a skilled artisan not only could have made but would have been motivated to make the combinations or modifications of prior art to arrive at the claimed invention.”). The Examiner also has not shown that the additional teachings of Joa, Li, and Gunawardana make up for the above-noted deficiencies of Zohar and Chang. As the Examiner has not shown that the asserted art combination teaches or suggests the disputed “overlap” limitation of claim 1, we do not sustain the Examiner’s obviousness rejection of independent claim 1, and claims 3–9 dependent therefrom. We also do not sustain the Examiner’s obviousness rejection of independent claims 10 and 17 (argued for substantially the same reasons as claim 1 and reciting similar limitations), and claims 12–16, 19, and 20 dependent therefrom. (Appeal Br. 18–19.) DECISION SUMMARY The Examiner’s rejection of claims 1, 3–10, 12–17, 19, and 20 under 35 U.S.C. § § 112(a) is REVERSED. The Examiner’s rejection of claims 1, 3–10, 12–17, 19, and 20 under 35 U.S.C. § 101 is AFFIRMED. The Examiner’s rejection of claims 1, 3–10, 12–17, 19, and 20 under 35 U.S.C. § 103 is REVERSED. Appeal 2020-000527 Application 14/616,543 28 In summary: Claims Rejected 35 U.S.C. § Reference(s)/ Basis Affirmed Reversed 1, 3–10, 12–17, 19, 20 112(a) Written Description 1, 3–10, 12–17, 19, 20 1, 3–10, 12–17, 19, 20 101 Eligibility 1, 3–10, 12–17, 19, 20 1, 8–10, 17 103 Zohar, Chang, Joa, Li 1, 8–10, 17 3–7, 12–16, 19, 20 103 Zohar, Chang, Joa, Li, Gunawardana 3–7, 12–16, 19, 20 Overall Outcome 1, 3–10, 12–17, 19, 20 Because we have affirmed at least one ground of rejection with respect to each claim on appeal, the Examiner’s decision is affirmed. See 37 C.F.R. § 41.50(a)(1). 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)(1)(iv). AFFIRMED Copy with citationCopy as parenthetical citation