DETAILED ACTION
Status of the Application
Claims 1-20 are pending and currently under consideration for patentability under 37 CFR 1.104.
Priority
The instant application has a filing date of October 4, 2023 and does not claim for the benefit of a prior-filed application.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claims 6, 12, and 17 are objected to because of the following informalities: --machine learning-- should be inserted preceding “model” in the phrase “the propensity model” to maintain consistency of terminology throughout the claims (the independent claims refer to a propensity machine learning model). Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads 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.
v Claim(s) 1-20 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Step 1:
Claim(s) 1-10 is/are drawn to methods (i.e., a process), claim(s) 11-15 is/are drawn to systems (i.e., a machine/manufacture), and claim(s) 16-20 is/are drawn to a non-transitory, computer-readable medium (i.e., a machine/manufacture). As such, claims 1-20 is/are drawn to one of the statutory categories of invention (Step 1: YES).
Step 2A - Prong One:
In prong one of step 2A, the claim(s) is/are analyzed to evaluate whether it/they recite(s) a judicial exception.
Claim 1 (representative of independent claim(s) 11 and 16) recites/describes the following steps;
receiving a set of features representing attributes corresponding to a first user of a digital platform;
receiving data for a plurality of digital campaigns, wherein data for each digital campaign in the plurality of digital campaigns indicates one or more digital components to send to user devices associated with users receiving that campaign;
processing the set of features…to determine whether the first user is expected to perform an affirmative action in response to the digital campaign, wherein the affirmative action includes performing one or more operations to transition the first user from a first user category to a second user category on the digital platform;
generating, using a feature importance model for each digital campaign in the plurality of digital campaigns, a subset of contributing features for the first user that are indicative of a likelihood of the first user performing the affirmative action in response to the digital campaign;
for each digital campaign, processing the respective subset of contributing features to generate a respective set of network inputs corresponding to the digital campaign, wherein each network input represents a user feature related to a particular digital campaign;
processing each set of network inputs using a respective trained model for each digital campaign in the plurality of digital campaigns, to obtain a respective output indicating a likelihood that the first user will perform an affirmative action in response to the respective digital campaign;
comparing the outputs from the trained student-teacher neural networks to identify a particular digital campaign in response to which the first user has a highest likelihood of performing the affirmative action; and
transmitting one or more components associated with the particular digital campaign to the first user
These steps, under its broadest reasonable interpretation, describe or set-forth a process for determining which advertising campaign content to send to a user (i.e., the one that has the highest likelihood of eliciting an affirmative action from the user) based on features/attributes of the user and components of the advertising campaigns. More specifically, the process comprises receiving a set of features representing attributes corresponding to a first user of a digital platform; receiving data for a plurality of digital campaigns, wherein data for each digital campaign in the plurality of digital campaigns indicates one or more digital components to send to user devices associated with users receiving that campaign; processing the set of features to determine whether the first user is expected to perform an affirmative action in response to the digital campaign, wherein the affirmative action includes performing one or more operations to transition the first user from a first user category to a second user category on the digital platform; generating, using a feature importance model for each digital campaign in the plurality of digital campaigns, a subset of contributing features for the first user that are indicative of a likelihood of the first user performing the affirmative action in response to the digital campaign; for each digital campaign, processing the respective subset of contributing features to generate a respective set of network inputs corresponding to the digital campaign, wherein each network input represents a user feature related to a particular digital campaign; processing each set of network inputs using a respective trained model for each digital campaign in the plurality of digital campaigns, to obtain a respective output indicating a likelihood that the first user will perform an affirmative action in response to the respective digital campaign; comparing the outputs from the trained student-teacher neural networks to identify a particular digital campaign in response to which the first user has a highest likelihood of performing the affirmative action; and transmitting one or more components associated with the particular digital campaign to the first user. This process amounts to a commercial or legal interactions (specifically, an advertising, marketing or sales activity or behavior; business relations). These limitations therefore fall within the “certain methods of organizing human activity” subject matter grouping of abstract ideas.
As such, the Examiner concludes that claim 1 recites an abstract idea (Step 2A – Prong One: YES).
Independent claim(s) 11 and 16 recite/describe nearly identical steps (and therefore also recite limitations that fall within this subject matter grouping of abstract ideas), and this/these claim(s) is/are therefore determined to recite an abstract idea under the same analysis.
Each of the depending claims likewise recite/describe these steps (by incorporation - and therefore also recite limitations that fall within this subject matter grouping of abstract ideas), and this/these claim(s) is/are therefore determined to recite an abstract idea under the same analysis. Any element(s) recited in a dependent claim that are not specifically identified/addressed by the Examiner under step 2A (prong two) or step 2B of this analysis shall be understood to be an additional part of the abstract idea recited by that particular claim. The same reasoning is similarly applicable to the limitations in the remaining dependent claims, and their respective limitations are not reproduced here for the sake of brevity.
Step 2A - Prong Two:
In prong two of step 2A, an evaluation is made whether a claim recites any additional element, or combination of additional elements, that integrate the exception into a practical application of that exception. An “addition element” is an element that is recited in the claim in addition to (beyond) the judicial exception (i.e., an element/limitation that sets forth an abstract idea is not an additional element). The phrase “integration into a practical application” is defined as requiring an additional element or a combination of additional elements in the claim to apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception, such that it is more than a drafting effort designed to monopolize the exception.
The claim(s) recite the additional elements/limitations of
“computer-implemented” (claim 1)
“a system comprising: at least one memory storing instructions; and at least one hardware processor interoperably coupled with the at least one memory, wherein execution of the instructions by the at least one hardware processor causes performance of operations” (claim 11)
“a non-transitory, computer-readable medium storing computer-readable instructions, that upon execution by at least one hardware processor, cause performance of operations” (claim 16)
“using a propensity machine learning model” (claims 1, 11, and 16)
“using a respective trained student-teacher neural network for each digital campaign…outputs from the trained student-teacher neural networks…” (claims 1, 11, and 16)
“transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user” (claims 1, 11, and 16)
“each student-teacher neural network model…using a trained teacher neural network of the student-teacher neural network…training a student neural network of the student-teacher neural network…wherein the student neural network processes…” (claim 2)
“a student neural network of the student-teacher neural network…output of the teacher neural network and a proposed output of the student neural network…output of the student neural network…” (claim 3)
“initial output of the teacher neural network and the proposed output of the student neural network…” (claim 4)
“output of the student neural network” (claim 5)
“wherein the propensity model is a gradient boosting model” (claims 6, 12, and 17)
“wherein the feature importance model is a gradient boosting model” (claims 7, 13, and 18)
“the student neural network and the teacher neural network” (claim 8)
“each trained student- teacher neural network” (claims 9, 14, and 20)
The requirement to execute the claimed steps/functions using “computer-implemented” means (claim 1) or “a system comprising: at least one memory storing instructions; and at least one hardware processor interoperably coupled with the at least one memory, wherein execution of the instructions by the at least one hardware processor causes performance of operations” (claim 11) or “a non-transitory, computer-readable medium storing computer-readable instructions, that upon execution by at least one hardware processor, cause performance of operations” (claim 16) is equivalent to adding the words “apply it” on a generic computer and/or mere instructions to implement the abstract idea on a generic computer. Applicant’s own disclosure explains that these “additional” elements may be embodied as a general-purpose computer (e.g., the published specification at paragraphs [0033]-[0047] & [0061] & [0074] & [0083]-[0093]). This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(f)).
The recitation of “using a propensity machine learning model” (claims 1, 11, and 16) and/or “using a respective trained student-teacher neural network for each digital campaign…outputs from the trained student-teacher neural networks…” (claims 1, 11, and 16) and/or “each student-teacher neural network model…using a trained teacher neural network of the student-teacher neural network…training a student neural network of the student-teacher neural network…wherein the student neural network processes…” (claim 2) and/or “a student neural network of the student-teacher neural network…output of the teacher neural network and a proposed output of the student neural network…output of the student neural network…” (claim 3) and/or “initial output of the teacher neural network and the proposed output of the student neural network…” (claim 4) and/or “output of the student neural network” (claim 5) and/or “wherein the propensity model is a gradient boosting model” (claims 6, 12, and 17) and/or “wherein the feature importance model is a gradient boosting model” (claims 7, 13, and 18) and/or the recitation of “the student neural network and the teacher neural network” (claim 8) and/or the recitation of “each trained student- teacher neural network” (claims 9, 14, and 20) provides nothing more than mere instructions to implement an abstract idea on a generic computer. See MPEP 2106.05(f) and the July 2024 Subject Matter Eligibility Examples and corresponding analysis. MPEP 2106.05(f) provides the following considerations for determining whether a claim simply recites a judicial exception with the words “apply it” (or an equivalent), such as mere instructions to implement an abstract idea on a computer: (1) whether the claim recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished; (2) whether the claim invokes computers or other machinery merely as a tool to perform an existing process; and (3) the particularity or generality of the application of the judicial exception. The “propensity machine learning model” (claims 1, 11, and 16) and “student-teacher neural networks…” (claims 1, 11, and 16) and and/or the description of “wherein the propensity model is a gradient boosting model” (claims 6, 12, and 17) and the description of “wherein the feature importance model is a gradient boosting model” (claims 7, 13, and 18) is used to generally apply the abstract idea without placing any limits on how the“propensity machine learning model” (or “gradient boosting model”) functions, without placing any limits on how the“student-teacher neural networks” (or associated student/teacher neural networks) function, and without placing any limits on how the“feature importance model” (gradient boosting model) functions. . Rather, these limitations only recite the outcomes of “to determine whether the first user is expected to perform an affirmative action in response to the digital campaign” (corresponding to the propensity machine learning model), and “generating…a subset of contributing features for the first user that are indicative of a likelihood of the first user performing the affirmative action in response to the digital campaign” (corresponding to the feature importance model), and “to obtain a respective output indicating a likelihood that the first user will perform an affirmative action in response to the respective digital campaign” (corresponding to the student-teacher neural networks), and do not include any details about how these outcome determinations are accomplished. See MPEP 2106.05(f) and the July 2024 Subject Matter Eligibility Examples and corresponding analysis. This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(f)).
The recited additional element(s) of “transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user” (claims 1, 11, and 16) serves merely to generally link the use of the judicial exception to a particular technological environment or field of use. Specifically, it/they serve(s) to limit the application of the abstract idea to computing environments, such as distributed computing environments and/or the internet, where information is represented digitally, exchanged between computers over a network, and presented using graphical user interfaces. It also serves merely to limit the advertising to digital forms of advertising/components as opposed to physical forms of advertising (e.g., printed mailers). This reasoning was demonstrated in Bilski, where it was determined that certain claim elements limiting the basic concept of hedging to commodities and energy markets (merely limiting an abstract idea to one field of use) did not make the concept patentable. This reasoning was demonstrated in Intellectual Ventures I LLC v. Capital One Bank (Fed. Cir. 2015), where the court determined "an abstract idea does not become nonabstract by limiting the invention to a particular field of use or technological environment, such as the Internet [or] a computer"). This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(g)).
The recitation of “using a propensity machine learning model” (claims 1, 11, and 16) and/or “using a respective trained student-teacher neural network for each digital campaign…outputs from the trained student-teacher neural networks…” (claims 1, 11, and 16) and/or “each student-teacher neural network model…using a trained teacher neural network of the student-teacher neural network…training a student neural network of the student-teacher neural network…wherein the student neural network processes…” (claim 2) and/or “a student neural network of the student-teacher neural network…output of the teacher neural network and a proposed output of the student neural network…output of the student neural network…” (claim 3) and/or “initial output of the teacher neural network and the proposed output of the student neural network…” (claim 4) and/or “output of the student neural network” (claim 5) and/or “wherein the propensity model is a gradient boosting model” (claims 6, 12, and 17) and/or “wherein the feature importance model is a gradient boosting model” (claims 7, 13, and 18) and/or the recitation of “the student neural network and the teacher neural network” (claim 8) and/or the recitation of “each trained student- teacher neural network” (claims 9, 14, and 20) also merely indicates a field of use or technological environment in which the judicial exception is performed. These types of limitations merely confine the use of the abstract idea to a particular technological environment (machine learning modes, student-teacher neural networks, gradient boosting model) and thus fails to add an inventive concept to the claims. In other words, Applicant is merely applying non-inventive machine-learning techniques/models, recited at a high level of generality, to process the associated data to generate the desired outputs. See MPEP 2106.05(h) and the July 2024 Subject Matter Eligibility Examples and corresponding analysis. This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(g)).
The recited additional element(s) of “transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user” also simply append insignificant extra-solution activity to the judicial exception, (e.g., mere post-solution activity in conjunction with an abstract idea). The term “extra-solution activity” is understood as activities incidental to the primary process or product that are merely a nominal or tangential addition to the claim. The recited additional element(s) do are deemed “extra-solution” because such data gathering and solution-outputting/transmission steps have long been held to be insignificant pre/post-solution activity. This/these limitation(s) do/does not impose any meaningful limits on practicing the abstract idea, and therefore do/does not integrate the abstract idea into a practical application (see MPEP 2106.05(h) and (g)).
Furthermore, although the claims recite a specific sequence of computer-implemented functions, and although the specification suggests certain functions may be advantageous for various reasons (e.g., business reasons), the Examiner has determined that the ordered combination of claim elements (i.e., the claims as a whole) are not directed to an improvement to computer functionality/capabilities, an improvement to a computer-related technology or technological environment, and do not amount to a technology-based solution to a technology-based problem. For example, Applicant’s published specification suggests that it is advantageous to implement the claimed business process for determining which advertising campaign content to send to a user because doing so can help improve the effectiveness of the advertiser’s marketing campaign by presenting advertising most likely to elicit a desired response (see, for example, Applicant’s published disclosure at paragraphs [0005] & [0015]). These are non-technical business advantages/improvements. At most, the ordered combination of claim elements is directed to a non-technical improvement to an abstract idea itself (e.g., a process for determining which advertising campaign content to send to a user). Examiner notes that although the specification asserts various improvements associated with computing resources, the Examiner is not persuaded that the various machine learning models or machine learning techniques (e.g., training of the student-teacher ANN models recited in claims 2-5) integrate the underlying judicial exception into a practical application. The Examiner finds the recitation of these various machine learning models/techniques to be analogous to those recited in the claims of Recentive Analytics, where the court explored the inquiry of whether claims that do no more than apply established methods of machine learning to a new data environment are patent eligible. The court held that they are not. Central to the courts determination of ineligibility was that “(b)oth sets of patents rely on the use of generic machine learning technology in carrying out the claimed methods for generating event schedules and network maps…The machine learning technology described in the patents is conventional…”. The court further found “that the machine learning model be "iteratively trained" or dynamically adjusted in the Machine Learning Training patents do not represent a technological improvement…[because] Iterative training using selected training material and dynamic adjustments based on real-time changes are incident to the very nature of machine learning”. The court also concluded that “the only thing the claims disclose about the use of machine learning is that machine learning is used in a new environment.” Significantly, the court further noted that “We have also held the application of existing technology to a novel database does not create patent eligibility… Stated differently, patents may be directed to abstract ideas where they disclose the use of an "already available [technology], with [its] already available basic functions, to use as [a] tool[ ] in executing the claimed process." SAP Am., 898 F.3d at 1169-70. We think those cases are equally applicable in the machine learning context.”. The Examiner finds Applicant has applied generic/known machine learning technology (e.g., certain types of machine learning models, generic student-teacher ANN training and mode use) to a new data environment (i.e., the user data and campaign data the present invention is concerned with).
Dependent claims 10, 15, and 20 fail to include any additional elements. In other words, each of the limitations/elements recited in respective dependent claims 10, 15, and 20 is/are further part of the abstract idea as identified by the Examiner for each respective dependent claim (i.e. they are part of the abstract idea recited in each respective claim). For example, claim 10 recites “wherein the subset of contributing features comprises the set of features”. This is an abstract limitation which further sets forth the abstract idea encompassed by claim 10. This limitation is not an “additional element”, and therefore it is not subject to further analysis under Step 2A- Prong Two or Step 2B. The same logic applies to each of the other dependent claims, whose limitations are not being repeated here for the sake of brevity and clarity. With respect to the other dependent claims not specifically listed here - each of the limitations/elements recited in these dependent claims other than those identified as being “additional” elements above (at the beginning of the Prong One analysis), are further part of the abstract idea encompassed by each respective dependent claim (i.e. it should be understood that these limitations are part of the abstract idea recited in each respective claim).
The Examiner has therefore determined that the additional elements, or combination of additional elements, do not integrate the abstract idea into a practical application. Accordingly, the claim(s) is/are directed to an abstract idea (Step 2A – Prong two: NO).
Step 2B:
In step 2B, the claims are analyzed to determine whether any additional element, or combination of additional elements, is/are sufficient to ensure that the claims amount to significantly more than the judicial exception. This analysis is also termed a search for an "inventive concept." An "inventive concept" is furnished by an element or combination of elements that is recited in the claim in addition to (beyond) the judicial exception, and is sufficient to ensure that the claim as a whole amounts to significantly more than the judicial exception itself. Alice Corp., 134 S. Ct. at 2355, 110 USPQ2d at 1981 (citing Mayo, 566 U.S. at 72-73, 101 USPQ2d at 1966)
As discussed above in “Step 2A – Prong 2”, the requirement to execute the claimed steps/functions using “computer-implemented” means (claim 1) or “a system comprising: at least one memory storing instructions; and at least one hardware processor interoperably coupled with the at least one memory, wherein execution of the instructions by the at least one hardware processor causes performance of operations” (claim 11) or “a non-transitory, computer-readable medium storing computer-readable instructions, that upon execution by at least one hardware processor, cause performance of operations” (claim 16) is equivalent to adding the words “apply it” on a generic computer and/or mere instructions to implement the abstract idea on a generic computer. These limitations therefore do not qualify as “significantly more” (see MPEP 2106.05(f)).
As discussed above in “Step 2A – Prong 2”, the recitation of “using a propensity machine learning model” (claims 1, 11, and 16) and/or “using a respective trained student-teacher neural network for each digital campaign…outputs from the trained student-teacher neural networks…” (claims 1, 11, and 16) and/or “each student-teacher neural network model…using a trained teacher neural network of the student-teacher neural network…training a student neural network of the student-teacher neural network…wherein the student neural network processes…” (claim 2) and/or “a student neural network of the student-teacher neural network…output of the teacher neural network and a proposed output of the student neural network…output of the student neural network…” (claim 3) and/or “initial output of the teacher neural network and the proposed output of the student neural network…” (claim 4) and/or “output of the student neural network” (claim 5) and/or “wherein the propensity model is a gradient boosting model” (claims 6, 12, and 17) and/or “wherein the feature importance model is a gradient boosting model” (claims 7, 13, and 18) and/or the recitation of “the student neural network and the teacher neural network” (claim 8) and/or the recitation of “each trained student- teacher neural network” (claims 9, 14, and 20) is equivalent to adding the words “apply it” on a generic computer and/or mere instructions to implement the abstract idea on a generic computer. These limitations therefore do not qualify as “significantly more” (see MPEP 2106.05(f)).
As discussed above in “Step 2A – Prong 2”, the recited additional element(s) of “transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user” (claims 1, 11, and 16) serves merely to generally link the use of the judicial exception to a particular technological environment or field of use. These limitations therefore do not qualify as “significantly more” (see MPEP 2106.05(g)).
As discussed above in “Step 2A – Prong 2”, the recitation of “using a propensity machine learning model” (claims 1, 11, and 16) and/or “using a respective trained student-teacher neural network for each digital campaign…outputs from the trained student-teacher neural networks…” (claims 1, 11, and 16) and/or “each student-teacher neural network model…using a trained teacher neural network of the student-teacher neural network…training a student neural network of the student-teacher neural network…wherein the student neural network processes…” (claim 2) and/or “a student neural network of the student-teacher neural network…output of the teacher neural network and a proposed output of the student neural network…output of the student neural network…” (claim 3) and/or “initial output of the teacher neural network and the proposed output of the student neural network…” (claim 4) and/or “output of the student neural network” (claim 5) and/or “wherein the propensity model is a gradient boosting model” (claims 6, 12, and 17) and/or “wherein the feature importance model is a gradient boosting model” (claims 7, 13, and 18) and/or the recitation of “the student neural network and the teacher neural network” (claim 8) and/or the recitation of “each trained student- teacher neural network” (claims 9, 14, and 20) also serves merely to generally link the use of the judicial exception to a particular technological environment or field of use. These limitations therefore do not qualify as “significantly more” (see MPEP 2106.05(g)).
As discussed above in “Step 2A – Prong 2”, the recited additional element(s) of “transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user” (claims 1, 11, and 16) also simply append insignificant extra-solution activity to the judicial exception, (e.g., mere post-solution activity in conjunction with an abstract idea). These additional element(s), taken individually or in combination, additionally amount to well-understood, routine and conventional activities previously known to the industry, specified at a high level of generality, appended to the judicial exception. These additional elements, taken individually or in combination, are well-understood, routine and conventional to those in the field of advertising. These limitations therefore do not qualify as “significantly more”. (see MPEP 2106.05(d)). This conclusion is based on a factual determination. The determination that receiving data/messages over a network is well-understood, routine, and conventional is supported by Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362; TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014), and MPEP 2106.05(d)(II), which note the well-understood, routine, conventional nature of receiving data/messages over a network. Furthermore, Examiner takes Official Notice that these steps were well-understood, routine, and conventional at the effective filing date of the claimed invention. Furthermore, the lack of technical detail/description in Applicant’s own specification provides implicit evidence that these steps were well-understood, routine, and conventional.
Viewing the additional limitations in combination also shows that they fail to ensure the claims amount to significantly more than the abstract idea. When considered as an ordered combination, the additional components of the claims add nothing that is not already present when considered separately, and thus simply append the abstract idea with words equivalent to “apply it” on a generic computer and/or mere instructions to implement the abstract idea on a generic computer, generally link the abstract idea to a particular technological environment or field of use, append the abstract idea with insignificant extra solution activity associated with the implementation of the judicial exception, (e.g., mere data gathering, post-solution activity), and appended with well-understood, routine and conventional activities previously known to the industry.
Dependent claims 10, 15, and 20 fail to include any additional elements. In other words, each of the limitations/elements recited in respective dependent claims 10, 15, and 20 is/are further part of the abstract idea as identified by the Examiner for each respective dependent claim (i.e. they are part of the abstract idea identified by the Examiner to which each respective claim is directed).
The Examiner has therefore determined that no additional element, or combination of additional claims elements is/are sufficient to ensure the claim(s) amount to significantly more than the abstract idea identified above (Step 2B: NO).
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(B) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention.
v Claims 1, 11, and 16 recite "processing the set of features using a propensity machine learning model to determine whether the first user is expected to perform an affirmative action in response to the digital campaign." There is insufficient antecedent basis for the “the digital campaign” limitation in this phrase. The claims previously refer to “data for a plurality of digital campaigns” and it is unclear as to which digital campaign is being referred. This is in contrast to subsequent references to “the digital campaign” in these claims, as subsequent references to “the digital campaign” all occur within descriptions of processes that are being performed for each digital campaign in the plurality. For the purpose of examination, the phrase “processing the set of features using a propensity machine learning model to determine whether the first user is expected to perform an affirmative action in response to the digital campaign” will be interpreted as being “processing the set of features using a propensity machine learning model to determine whether the first user is expected to perform an affirmative action in response to each digital campaign.”
Each of the dependent claims are similarly rejected by virtue of their dependency on one of these claims.
v Claims 1, 11, and 16 recite "transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user." There is insufficient antecedent basis for the “the particular digital campaign” limitation in this phrase. The claims previously recite “for each digital campaign…generate a respective set of network inputs corresponding to the digital campaign, wherein each network input represents a user feature related to a particular digital campaign” and later recite “comparing the outputs from the trained student-teacher neural networks to identify a particular digital campaign in response to which the first user has a highest likelihood of performing the affirmative action” and it is unclear as to which particular digital campaign is being referred. For the purpose of examination, the phrase “transmitting one or more digital components associated with the particular digital campaign to a user device associated with the first user” will be interpreted as being “transmitting one or more digital components associated with the identified particular digital campaign to a user device associated with the first user.”
Each of the dependent claims are similarly rejected by virtue of their dependency on one of these claims.
v Claims 1, 11, and 16 require “processing each set of network inputs using a respective trained student-teacher neural network for each digital campaign in the plurality of digital campaigns, to obtain a respective output indicating a likelihood that the first user will perform an affirmative action in response to the respective digital campaign; comparing the outputs from the trained student-teacher neural networks to identify a particular digital campaign in response to which the first user has a highest likelihood of performing the affirmative action” and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. The metes and bounds of the claimed “student-teacher neural network” are unclear, as it is unclear how and/or whether this element is different from the student neural network referred to in claim 2. As is known in the art, teacher-student training is a knowledge distillation framework/process for creating a student model based on a teacher model (i.e., distillation of a larger model into a smaller model). In other words, there is no single “student-teacher neural network” that is created using a teacher-student training framework. Claim 2 demonstrates this, at it describes using a teacher-student training framework to train a student neural network based on a teacher neural network. However, Applicant’s claims appear to suggest existence of a single “student-teacher neural network” that has/comprises both a trained teacher neural network and a trained student neural network (see claims 2 and 3). It is unclear how a single neural network would/could comprise both of these discrete neural networks, or how the network requirements are required to be processed using each single student-teacher neural network given each single student-teacher neural network apparently comprises multiple different neural networks. Therefore, the claim is indefinite for failing to particularly and distinctly claim the subject matter which the application regards as the invention.
Each of the dependent claims are similarly rejected by virtue of their dependency on one of these claims.
For the purpose of examination, the phrase “processing each set of network inputs using a respective trained student-teacher neural network for each digital campaign in the plurality of digital campaigns, to obtain a respective output indicating a likelihood that the first user will perform an affirmative action in response to the respective digital campaign; comparing the outputs from the trained student-teacher neural networks to identify a particular digital campaign in response to which the first user has a highest likelihood of performing the affirmative action” will be interpreted as being “processing each set of network inputs using a respective trained student neural network for each digital campaign in the plurality of digital campaigns, to obtain a respective output indicating a likelihood that the first user will perform an affirmative action in response to the respective digital campaign; comparing the outputs from the trained student neural networks to identify a particular digital campaign in response to which the first user has a highest likelihood of performing the affirmative action.”
Relatedly, for the purpose of examination, the recitation in claim 2 of “training each student-teacher neural network model, wherein the training comprises: receiving a set of unlabeled training data from the feature importance model; receiving, for a first set of unlabeled training data, a set of respective labels, wherein each label is associated with a contributing feature for the user that is indicative of a likelihood that the first user will respond affirmatively to the respective digital campaign; processing (1) the first set of unlabeled training data that is associated with the received set of labels and (2) a second set of unlabeled training data, to generate a set of network inputs, wherein each network input represents a contributing feature for the user; processing the set of network inputs using a trained teacher neural network of the student-teacher neural network that generates a respective initial output for each network input; and training a student neural network of the student-teacher neural network to optimize a loss function, wherein the student neural network processes the set of network inputs and generates a respective output for each network input” will be interpreted as being “training each student neural network model, wherein the training comprises: receiving a set of unlabeled training data from the feature importance model; receiving, for a first set of unlabeled training data, a set of respective labels, wherein each label is associated with a contributing feature for the user that is indicative of a likelihood that the first user will respond affirmatively to the respective digital campaign; processing (1) the first set of unlabeled training data that is associated with the received set of labels and (2) a second set of unlabeled training data, to generate a set of network inputs, wherein each network input represents a contributing feature for the user; processing the set of network inputs using a trained teacher neural network t.”
Relatedly, for the purpose of examination, the recitation in claim 3 of “wherein training a student neural network of the student-teacher neural network to optimize a loss function comprises, for each contributing feature for the user” will interpreted as “wherein training a student neural network ”.
Indication of Novel and Non-Obvious Subject Matter
Independent claims 1, 11, and 16 recite novel and non-obvious subject matter. Each of the dependent claims similarly recite novel and non-obvious subject matter by virtue of their dependency on one of these claims.
The following is an examiner’s statement of reasons for indication of novel and non-obvious subject matter:
The closest prior art of record is Chang et al. (U.S. PG Pub No. 2021/0374562 , December 2, 2021- hereinafter "Chang”); Ma et al. (U.S. PG Pub No. 2022/0358347 , November 10, 2022- hereinafter "Ma”); Aarabi (U.S. PG Pub No. 2024/0281660 , August 22, 2024- hereinafter "Aarabi”); Shamir et al. (U.S. PG Pub No. 2025/0077934 , March 6, 2025- hereinafter "Shamir”); Pham et al. (U.S. PG Pub No. 2022/0188636 , June 16, 2022- hereinafter "Pham”); Liu et al. (U.S. PG Pub No. 2023/0214670 , July 6, 2023- hereinafter "Liu”); Bent III et al. (U.S. PG Pub No. 2024/0126576 , April 18, 2024- hereinafter "Bent”); Yao et al. (U.S. PG Pub No. 2025/0252318 , August 7, 2025- hereinafter "Yao”); Li et al. (Chinese. PG Pub No. CN 115271272 A , November 1, 2022 - hereinafter "Li”); Zhao (Chinese PG Pub No. CN 11610866 A , May 12, 2023- hereinafter "Zhao”); Wang et al. (U.S. PG Pub No. 2019/0325923 , October 24, 2019- hereinafter "Wang”); “Distilled CTR: Accurate and scalable CTR prediction model through model distillation” (Jose, Alijo and Shetty, Sujala; published January 3, 2022 https://doi.org/10.1016/j.eswa.2021.116474).
Chang discloses a process for ranking for content (e.g., advertisements, product recommendations) based on scores (e.g., representing probability of the user performing a certain action such as clicking/converting on the content) using a compressed model (using knowledge distillation) based on a large baseline version of a prediction ML model (e.g., neural network) and further based on a reduced set of identified high-importance features (i.e., those that determined to have a high impact on the probability of interaction with the content item). Ma discloses use of a feature extraction/importance model to identify important features of a user/content that are correlated with historical interest with content (e.g., advertisement, recommendation), training of a large neural network trained to predict engagement with content items (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large model into a smaller student model (e.g., to minimize a loss term that measures mean squared error between model outputs, wherein the loss term is a cross entropy loss), and using the student model to score candidate content for probability of interaction/engagement, and transmitting the highest rank content to a user.
Aarabi discloses extracting sets of features from content, training of a large neural network trained to predict engagement with content items (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large model into a smaller student model, using the student model to generate click/selection probabilities for different advertisements, and selecting an advertisement having the highest probability to serve to a user.
Shamir discloses training of a deep neural network trained to predict engagement with content items (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large model into a smaller student model (e.g., to minimize a loss term that measures mean squared error between model outputs, wherein the loss term is a cross entropy loss), and using the student model to score candidate content for probability of interaction/engagement, and transmitting the highest rank content to a user.
Pham discloses extracting sets of user and content features (e.g., ad features, context features), training of a large neural network trained to predict engagement with content items (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large neural network into a smaller student neural network (e.g., to minimize a loss term that measures error between model outputs, wherein the loss term is a cross entropy loss), using the student model to generate click/selection probabilities for different advertisements, and selecting an advertisement having the highest probability to serve to a user.
Liu discloses using MAB to determine important user/content/context features associated with user engagement with content (e.g., advertisements), using a teacher-student training framework to distill a large neural network into a smaller student neural network (e.g., to minimize a loss term that measures error between model outputs, wherein the loss term is a cross entropy loss), and using the student model to score candidate content for probability of interaction/engagement, and transmitting the highest rank content to a user.
Bent discloses using a teacher-student training framework to distill a large neural network (e.g., pre-trained LLM, deep neural network) into a smaller student neural network (e.g., to minimize a loss term that measures mean squared error between model outputs, wherein the loss term is a cross entropy loss), and use of the student model to generate advertisements and/or discrete advertisement components having a high likelihood of eliciting a desired response. Discloses training respective student models for different types of content items (e.g., headlines, descriptions, images, etc.) for the advertisements.
Yao discloses use of a feature extraction/importance model to identify important features that are correlated predictive outputs, training one or more large neural network trained to make a prediction (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill one or more teacher neural networks into a smaller student neural networks (e.g., to minimize a loss term that measures error between model outputs, wherein the loss term is a cross entropy loss), merging the student and teacher neural networks to create a merged network, and using the merged network to make predictions.
Li discloses extracting sets of user and content features (e.g., ad features, context features) correlated with user clicks, training of a large neural network trained to predict engagement with content items (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large neural network into a smaller student neural network (e.g., to minimize a loss term that measures error between model outputs), using the student model to generate click/selection probabilities for different advertisements, and selecting an advertisement having the highest probability to serve to a user.
Zhao discloses extracting sets of user and content features (e.g., ad features, context features), training of a large neural network trained to predict clicks/conversions with advertisements (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large neural network into a smaller student neural network (e.g., to minimize a loss term that measures error between model outputs, wherein the loss term is a cross entropy loss), using the student model to generate click/conversion probabilities for different advertisements for a user.
Wang discloses using a gradient boosting model to generate a subset of important user features and advertising features that are highly correlated with clicks/conversions.
“Distilled CTR: Accurate and scalable CTR prediction model through model distillation” discloses training of a large neural network trained to predict clicks/conversions with advertisements (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large neural network into a smaller student neural network (e.g., to minimize a loss term that measures mean squared error between model outputs, wherein the loss term is a cross entropy loss), using the student model to generate click/conversion probabilities for different advertisements for a user.
As per claims 1, 11, and 16, the closest prior art of record taken either individually or in combination with other prior art of record fails to teach or suggest the specific data processing flow required by the claim language, with specific emphasis on the initial requirement to use a propensity machine learning model to process the set of features to determine whether the first user is expected to perform one or more operations to transition the first user from a first user category to a second user category on the digital platform in response to the digital campaign, in combination with the subsequent requirement to use a respective feature importance model for each campaign to generate a subset of contributing features for the first user for each of the plurality of campaigns, and using a respective trained student-teacher neural network for each campaign of the plurality of campaigns to process each respective set of network inputs to obtain respective likelihoods that the first user will perform an affirmative action in response to each respective campaign (i.e., a respectively-trained student model for each campaign). Although the prior art discloses use of a propensity machine learning model to process a set of user features to determine whether the first user is expected to perform one or more operations to transition the first user from a first user category to a second user category on the digital platform in response to a digital campaign, using a gradient boosting model to generate a subset of important user features and advertising features that are highly correlated with clicks/conversions, training of a large neural network trained to predict engagement with content items (e.g., using a large set of unlabeled data as well as respective labels), using a teacher-student training framework to distill the large model into a smaller student model (e.g., to minimize a loss term that measures mean squared error between model outputs, wherein the loss term is a cross entropy loss – using the training process recited in dependent claims 2-5), and using the student model to score candidate content for probability of interaction/engagement, and transmitting the highest rank content to a user – the specific data processing flow required by the claim language and the combination of features emphasized above would not have been obvious absent applicants’ own disclosure other than with impermissible hindsight.
Claims 2-10, 12-15, and 17-20 depend upon claims 1, 11, or 16 and have all the limitations of claims 1, 11, or 16, and therefore similarly recite novel and non-obvious subject matter.
Conclusion
No claim is allowed
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JAMES M DETWEILER whose telephone number is (571)272-4704. The examiner can normally be reached on Monday-Friday from 8 AM to 5 PM ET.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Waseem Ashraf can be reached at telephone number (571)-270-3948. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center for authorized users only. Should you have questions about access to Patent Center, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) Form at https://www.uspto.gov/patents/uspto-automated- interview-request-air-form.
/JAMES M DETWEILER/Primary Examiner, Art Unit 3621
Prosecution Insights