Lebrun, AlexandreDownload PDFPatent Trials and Appeals BoardSep 4, 201914686771 - (R) (P.T.A.B. Sep. 4, 2019) 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. 14/686,771 04/14/2015 Alexandre Lebrun 26295-29145/US 5058 87851 7590 09/04/2019 Facebook/Fenwick Silicon Valley Center 801 California Street Mountain View, CA 94041 EXAMINER LELAND III, EDWIN S ART UNIT PAPER NUMBER 2677 NOTIFICATION DATE DELIVERY MODE 09/04/2019 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): fwfacebookpatents@fenwick.com ptoc@fenwick.com PTOL-90A (Rev. 04/07) UNITED STATES PATENT AND TRADEMARK OFFICE ____________ BEFORE THE PATENT TRIAL AND APPEAL BOARD ____________ Ex parte ALEXANDRE LEBRUN ___________ Appeal 2018-005801 Application 14/686,771 Technology Center 2600 _________________ Before JASON V. MORGAN, JAMES B. ARPIN, and NABEEL U. KHAN, Administrative Patent Judges. ARPIN, Administrative Patent Judge. DECISION ON REQUEST FOR REHEARING Pursuant to 37 C.F.R. § 41.52, Appellant1 filed a Request for Rehearing (“Req. Reh’g”), dated August 5, 2019, seeking reconsideration of our Decision on Appeal mailed June 4, 2019 (‘Dec.’),2 in which we affirmed the Examiner’s rejection of claims 1–15 under 35 U.S.C. § 101 as directed to 1 Appellant asserts Facebook, Inc., is the real party-in-interest. App. Br. 2. 2 In this Decision, we refer to Appellant’s Appeal Brief (“App. Br.,” filed January 29, 2018) and Reply Brief (“Reply Br.,” filed May 18, 2018); the Final Office Action (“Final Act.,” mailed July 12, 2017); the Examiner’s Answer (“Ans.,” mailed March 19, 2018; and the originally-filed Specification (“Spec.,” filed April 14, 2015). Rather than repeat the Examiner’s findings and determinations and Appellant’s contentions in their entirety, we refer to these documents. Appeal 2018-005801 Application 14/686,771 2 patent-ineligible subject matter. Final Act. 3–7. We have jurisdiction over the Request for Rehearing under 35 U.S.C. § 6(b). STATEMENT OF THE CASE Appellant’s claimed methods and computer-readable media “relate[] in general to the field of natural language processing, and in particular to an email-like user interface and a crowd-source based network for configuring and training a natural language system interfaced with a runtime system or application.” Spec. ¶ 2 (emphasis added). As noted above, claims 1–15 are pending. Claims 1 and 9 are independent. App. Br. 12–13 (claim 1), 14–15 (claim 9) (Claims App.). Claims 2–8 depend directly or indirectly from claim 1, and claims 10–15 depend directly or indirectly from claim 9. Id. at 12–16. Claim 1, reproduced below, is representative. 1. A computer-implemented method comprising: maintaining, by a configuration system, a plurality of natural language (NL) instances, each NL instance developed by one or more developers and comprising 1) a set of training data and 2) a prediction model that predicts a user-desired application function based on a NL query, the set of training data being used to train the prediction model of each NL instance by mapping application functions to NL queries; receiving, by the configuration system, one or more linking requests from the one or more developers to link NL instances to other NL instances of the plurality of NL instances, each of the one or more developers in communication with the configuration system via a network and at least one of the one or more developers being separate from the configuration and other developers of the one or more developers; linking, by the configuration system, one or more NL instances of the plurality of NL instances to one or more other Appeal 2018-005801 Application 14/686,771 3 NL instances of the plurality of NL instances responsive to the received linking requests; and training, by the configuration system, each of the NL instances of the plurality of NL instances using: (1) the set of training data for the NL instance, and (2) each set of training data for other NL instances to which the NL instance is linked; for each of one or more of the plurality of NL instances: receiving, from a client device of a user, an NL query corresponding to an application request for the NL instance to provide a user-desired application function based on the NL query, using the prediction model for the NL instance to determine an application function according to the prediction model, and responding to the NL query of the user by providing the determined application function to the client device. Id. at 12–13. “The developer platform’s inbox allows for minimizing the number of human interactions required to improve the quality and accuracy of an NL configuration system in processing natural language expressions.” Spec. ¶ 5; see id. ¶¶ 6, 8; see also Reply Br. 3 (“[T]the claimed method allows a Natural Language (NL) configuration system to provide a collaboration platform where developers configure and optimize their natural language systems by leveraging the work and prediction model training data of other developers when generating and configuring their own prediction models.” (emphases added)). Because Appellant only argues the limitations of independent claim 1, we understand that claims 2–15 stand or fall with claim 1. For the reasons given below, we continue to sustain the Examiner’s rejection. Appeal 2018-005801 Application 14/686,771 4 DISCUSSION 1. Overview Appellant asserts that (1) we mistakenly determined that claim 1 recited “[c]ertain methods of organizing human activity,” i.e., an abstract idea, in view of the “wake-up call” example described in the Specification (Spec. ¶ 19); and (2) we overlooked claim 1’s similarity to Example 39 from the Subject Matter Eligibility Examples that accompanied the 2019 Revised Patent Subject Matter Eligibility Guidance (2019 Subject Matter Eligibility Examples: Abstract Ideas, 8–9 (Jan. 7, 2019)). Req. Reh’g 2–6. Initially, we note that, although the “wake-up call” example appears in the Specification, Appellant did not raise this example in the Appeal or Reply Brief. Further, although Appellant asserted the similarity of the pending claims to Example 1 of the Examples: Abstract Ideas,3 accompanying the 2014 Interim Eligibility Guidance and identifying a hypothetical patent eligible claim, in the Appeal and Reply Briefs (App. Br. 8–9; Reply Br. 2), Example 39 was not available when the Examiner mailed the patent eligibility rejection to the pending claims or when Appellant filed the Appeal or Reply Brief or when the Examiner mailed the Answer. See supra, note 2. Consequently, Appellant did not raise arguments regarding Example 39 before this Request. Generally, “[a]rguments not raised, and Evidence not previously relied upon, pursuant to §§ 41.37, 41.41, or 41.47 are not permitted in the request for rehearing.” 37 C.F.R. § 41.52(a)(1) (emphasis added). Moreover, the Office’s 2019 Revised Patent Subject Matter Eligibility 3 App. Br. 8 (citing https://www.uspto.gov/sites/default/files/documents/ abstract_idea_examples.pdf). Appeal 2018-005801 Application 14/686,771 5 Guidance is not a change in the law, but, instead, instructs Office personnel on how to consistently and predictably apply the U.S. Supreme Court’s two- part test for determining patent eligibility.4 Further, we did not raise a new ground of rejection. See Dec. 1, 18 (affirming the Examiner’s rejection). Thus, neither Appellant’s contentions relying upon the “wake-up call” example nor those relying upon Example 39 satisfy the exceptions to the requirements for requests for rehearing. See 37 C.F.R. § 41.52(a)(2)–(4). Nevertheless, on this record and given the timing of the release of the guidance and examples, we exercise our discretion here to consider Appellant’s new contentions,5 but, for the reasons given below, we deny Appellant’s request that we reverse the Examiner’s rejection of claims 1–15. 4 As the Office explains: This guidance does not constitute substantive rulemaking and does not have the force and effect of law. The guidance sets out agency policy with respect to the USPTO’s interpretation of the subject matter eligibility requirements of 35 U.S.C. 101 in view of decisions by the Supreme Court and the Federal Circuit. The guidance was developed as a tool for internal USPTO management and does not create any right or benefit, substantive or procedural, enforceable by any party against the USPTO. Rejections will continue to be based upon the substantive law, and it is those rejections that are appealable to the Patent Trial and Appeal Board (PTAB) and the courts. 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50, 51 (Jan. 7, 2019) (emphasis added). 5 Our exercise of discretion here deviates somewhat from the Office’s preference for new contentions to be presented first to the examiner, so that the appellant(s) and the examiner may address the merits of new contentions during the normal course of prosecution. See 37 C.F.R. § 41.37(c)(1)(iv); but see MPEP § 1205.02 (“This sentence is not intended to preclude the filing of a supplemental brief or document if new authority should become available or relevant after the brief or reply brief was filed. An example of Appeal 2018-005801 Application 14/686,771 6 2. “Certain methods of organizing human activity” Appellant contends that the claims “that recite training a computer model do not recite a judicial exception under Step 2A, Prong 1.” Req. Reh’g 2. Under the Step 2A, Prong One, of the Office’s 2019 Revised Patent Subject Matter Eligibility Guidance, if the claims are directed to a recognized statutory category, we then look to whether the claims recite any judicial exceptions, including certain groupings of abstract ideas (i.e., mathematical concepts, certain methods of organizing human activity, such as a fundamental economic practice, or mental processes). See 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. at 54–55 (“Revised Step 2A”). In our Decision, we determined claim 1’s limitations, under their broadest reasonable interpretation, recite “managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules or instructions),” which falls within the broader category of “[c]ertain methods of organizing human activity.” Dec. 10 (quoting 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. at 52). Thus, we determined that the rejected claims recite an abstract idea, namely “[c]ertain methods of organizing human activity.” Id. at 11; see Final Act. 3, 6; Ans. 4–6; 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. at 52. Appellant contends: While the claimed method enables a developer to use the training data from the work of other developers, the claim as a whole is not directed to managing relationships or interactions between people (e.g., developers). Specifically, the claimed linking is not such circumstances would be where a pertinent decision of a court or other tribunal was not published until after the brief or reply brief was filed.”). Appeal 2018-005801 Application 14/686,771 7 a way to connect people, like developers. Instead, the claimed linking is a way for a developer to define which training data will be used to train the developer’s prediction model. Req. Reh’g 2–3. As an example of the recited method, Appellant asserts the prediction model may receive the NL query, “Wake me up at 6 tomorrow morning,” and the model would then output the function “setalarm(0600)” to an alarm application, thereby causing the alarm application to perform the application function of creating a new alarm at 6 am. See Spec. at ¶ 19. The prediction model and a set of training data used to train the prediction model form an “NL instance.” Id. at 3. We do not find this example sufficient to show that we were mistaken in determining that the claim recites an abstract idea. If a person requests (e.g., “receiving . . . an NL query corresponding to an application request for the NL instance to provide a user-desired application function based on the NL query”) another person (e.g., a developer) to wake her up at 6:00am (e.g., an interaction between persons involving a “natural language expression”6), the other person can take in her surroundings to determine what options for satisfying the request are available (e.g., gather “a set of training data” forming a portion of an NL instance) and assess the possible 6 The Specification explains that: The user inputs the NL query on the runtime system 112 through voice commands, touch-screen gestures, audio signals or keystrokes. An explicit form of an NL query includes as a natural language expression. Example expressions of a NL query for setting an alarm may include: “Wake me up at 6 tomorrow morning,” “Set the alarm at 6 am,” and “I’d like you to wake me up by 6 in the morning.” Spec. ¶ 19; see id. ¶ 3. Appeal 2018-005801 Application 14/686,771 8 replies satisfying the request (e.g., “linking” NL instances and “using the prediction model for the NL instance to determine an application function according to the prediction model”). If the requestor and the other person are in a hotel, the other person can call the front desk, which may be staffed or automated (e.g., another developer or a client device), and arrange for a wake-up call (e.g., a further interaction between persons); or the other person can set an alarm clock (i.e., “responding to the NL query of the user by providing the determined application function to the client device”) to sound a 6:00am alarm, or the other person can use the alarm application on the requester’s mobile phone (i.e., “responding to the NL query of the user by providing the determined application function to the client device”) to sound at 6:00am; or if the other person is getting up before 6:00am, the other person can interact directly with the requester and simply wake up the requester at 6:00am. Other options are possible. The decision on how to ensure timely wake-up may occur in the other person’s mind (e.g., the prediction model).7 Dec. 10; see Spec. ¶ 6 (“The NL configuration systems are domain agnostic and based on actual end-users’ interactions validated by each developer in the community for their own NL configuration system, thus minimizing the time required to bootstrap and improve such a system.”). 7 Although, in our Decision, we did not determine that the claims recited “[m]ental processes—concepts performed in the human mind (including an observation, evaluation, judgment, opinion),” if prosecution resumes, Appellant and the Examiner may wish to consider whether the claims also recite a mental process. See 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. at 52. Appeal 2018-005801 Application 14/686,771 9 This example clearly encompasses interactions between people, gathering and using of stored data, and invoking technology. See Final Act. 3–4, 7; Reply Br. 6; see also Dec. 8–10 (“Thus, we agree with the Examiner that the focus of this claim is on collecting information, modifying it, analyzing it, and displaying certain results of the collection, modification, and analysis. Ans. 6; see Electric Power Grp. Grp.[, LLC v. Alstrom S.A.], 830 F.3d [1350,] 1353 [(Fed. Cir. 2016)].”); Cellspin Soft, Inc. v. Fitbit, Inc., 927 F.3d 1306, 1315 (Fed. Cir. 2019) (“As the district court recognized, we have consistently held that similar claims reciting the collection, transfer, and publishing of data are directed to an abstract idea.”). Although the wake-up call example involves, and the claims recite, the production of “an application function” based on the NL query to reply to the query, this is merely the technological environment in which the reply to the NL query is formulated. See Dec. 13–14; MPEP § 2106.05(h). Thus, we do not find Appellant’s contention persuasive. 3. Hypothetical Claim of Example 39 Appellant further contends that Example 39 from the Subject Matter Eligibility Examples provides an exemplary claim for a method for training a machine learning model and that the patent eligible method of Example 39 parallels the methods recited in the rejected claims. Req. Reh’g 4. Appellant concludes that the reasoning provided with respect to Example 39 directly applies to the rejected claims. Id. “The Board decides cases in accordance with the law, not in accordance with hypothetical” examples. Ex Parte Blythe, Appeal No. 2017-003176, 2018 WL 3047568, at *8 (PTAB May 31, 2018) (nonprecedential). Example 39, like the other examples provided by the Appeal 2018-005801 Application 14/686,771 10 Office, is merely a “hypothetical [that is] only intended to be illustrative of the claim analysis under the 2019 [Patent Eligibility Guidance].” Subject Matter Eligibility Examples: Abstract Ideas, 1, available at https://www.uspto.gov/sites/default/files/documents/101_examples_37to42_ 20190107.pdf (Jan. 7, 2019). Nonetheless, we assess to what extent the fact pattern set forth by Example 39 bears on the claims of Appellant’s pending application. The hypothetical claim of Example 39 recites: A computer-implemented method of training a neural network for facial detection comprising: collecting a set of digital facial images from a database; applying one or more transformations to each digital facial image including mirroring, rotating, smoothing, or contrast reduction to create a modified set of digital facial images; creating a first training set comprising the collected set of digital facial images, the modified set of digital facial images, and a set of digital non-facial images; training the neural network in a first stage using the first training set; creating a second training set for a second stage of training comprising the first training set and digital non-facial images that are incorrectly detected as facial images after the first stage of training; and training the neural network in a second stage using the second training set. 2019 Subject Matter Eligibility Examples: Abstract Ideas, 8–9. Appellant contends “Example 39 recites a method for training a model, where the method obtains two different training sets and then broadly recites ‘training’ the model using those training sets.” Req. Reh’g 4. Appellant asserts pending claim 1 recites training a prediction model of an Appeal 2018-005801 Application 14/686,771 11 NL instance using: “(1) the set of training data for the NL instance, and (2) each set of training data for other NL instances to which the NL instance is linked.” App. Br. 12 (Claims App’x). Appellant concludes [j]ust as in Example 39, the claims recite how to obtain two sets of training data (e.g., one set that is part of the NL instance and another set that is part of another NL instance that was linked to) and then recite using those two sets of training data to train a model. Req. Reh’g 5. Appellant notes that, in finding the claims recite “[c]ertain methods of organizing human activity,” we reasoned that “claim 1 does not restrict the nature of the natural language instance prediction models, and, therefore, claim 1 does not preclude the steps directed to training the natural language instances and using them to determine an application function from encompassing ‘steps people go through in their minds, or by mathematical algorithms, without more.’” Id. (quoting Dec. 10). Appellant contends, however, that the patent eligible claim of Example 39 is no more restrictive than pending claim 1. Further, Appellant contends that we improperly discounted “the technical nature of the recited ‘training’ of the computer model because it ‘does not specify how “training” is accomplished, but instead focuses on what is used to perform the “training” step.’” Id. at 6 (citing Dec. 9 n.5). In particular, Appellant asserts, training a prediction model using the “the set of training data for the NL instance” and “each set of training data for other NL instances to which the NL instance is linked” is no more able to be “practically performed in the human mind” nor is it “managing interactions between people” than the claim in Example 39. Appeal 2018-005801 Application 14/686,771 12 Id.; see also 2019 Subject Matter Eligibility Examples: Abstract Ideas, 9. As noted above, in the 2019 Subject Matter Eligibility Examples: Abstract Ideas, the Office explains that the examples contained therein “are hypothetical and only intended to be illustrative of the claim analysis under the 2019 PEG. These examples should be interpreted based on the fact patterns set forth below as other fact patterns may have different eligibility outcomes.” 2019 Subject Matter Eligibility Examples: Abstract Ideas, 1 (emphasis added). Thus, we evaluate the pending claims in view of their recitations, as understood in light of the disclosure of the Specification, pursuant to the guidance provided by the Supreme Court and the Federal Circuit, as interpreted by the Office. Trading Techs. Int’l, Inc. v. IBG LLC, 921 F.3d. 1084, 1095 (Fed. Cir. 2019) (“We are not bound by non- precedential decisions at all, much less ones to different patents, different specifications, or different claims. Each panel must evaluate the claims presented to it. Eligibility depends on what is claimed, not all that is disclosed in the specification.”); see 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. at 52 n.13 (citing, e.g., Interval Licensing LLC v. AOL, Inc., 896 F.3d 1335, 1344–45 (Fed. Cir. 2018); Voter Verified, Inc. v. Election Systems & Software, LLC, 887 F.3d 1376, 1385 (Fed. Cir. 2018)). Mere similarity between the language of an hypothetical claim and a pending claim is not sufficient to ensure patent eligibility. See Cleveland Clinic Foundation v. True Health Diagnostics LLC, 760 Fed. Appx. 1013, 1021 (Fed. Cir. 2019) (“We have considered Example 29 and the arguments relating to it, but to the extent that Example 29–Claim 1 is analogous to the claims at issue, Ariosa [Diagnostics, Inc. v. Sequenom, Inc., 788 F.3d 1371, 1376–78 (Fed. Cir. 2015)] must control.”). Appeal 2018-005801 Application 14/686,771 13 As the Federal Circuit noted, we evaluate the patent eligibility of the claims presented to us based on what is recited in those claims, as understood in light of the Specification. See Trading Techs., 921 F.3d. at 1095. As we noted in our Decision: Claim 1 does not restrict the nature of the natural language instance prediction models, and, therefore, claim 1 does not preclude the steps directed to training the natural language instances and using them to determine an application function from encompassing “steps people go through in their minds, or by mathematical algorithms, without more.” Dec. 10 (citing Electric Power Grp., 830 F.3d at 1354). Appellant does not dispute this determination, but, instead, merely argues that the model recitations of the hypothetical claim of Example 39 are equally or more broad. Req. Reh’g 5–6. Because the Specification describes the “prediction model” broadly by what it does, rather than what it is, we understand the term “prediction model” broadly to encompass a suitable component “capable of predicting a user-desired application function.” Spec. ¶ 34; see id. ¶¶ 6, 28, 43. Unlike pending claim 1, the hypothetical claim of Example 39 recites how the neural network is trained and what it is trained to do in a manner that excludes mental steps alone. 2019 Subject Matter Eligibility Examples: Abstract Ideas, 8–9 (“applying one or more transformations to each digital facial image including mirroring, rotating, smoothing, or contrast reduction to create a modified set of digital facial images; creating a first training set comprising the collected set of digital facial images, the modified set of digital facial images, and a set of digital non-facial images; [and] creating a second training set for a second stage of training comprising the first training set and digital non-facial images that are incorrectly detected as facial images after Appeal 2018-005801 Application 14/686,771 14 the first stage of training” (emphasis added)); see id. at 8 (Background). Consequently, we conclude that the model recitations of the hypothetical claim of Example 39 are more specific and technical than those of the pending claims and contribute to the assessment of its eligibility and that pending claim 1 and the hypothetical claim of Example 39 are distinguishable. Appellant does not persuade us otherwise. For the reasons given above, we are not persuaded that we overlooked or misapprehended points of law or fact, such that we should reverse the Examiner’s determination that claims 1–15 are unpatentable. DECISION Appellant’s Request for Rehearing has been granted to the extent that we have reconsidered our Decision in light of Appellant’s Request, but is denied in all other respects. DENIED Copy with citationCopy as parenthetical citation