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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 1/16/26 has been entered.
Response to Amendment
Claims 1, 2, 5, 10, 11, 14, and 19 have been amended.
No claims have been added. Claims 8 and 17 have been canceled, with claims 3, 6-7, 12, 15-16 being previously canceled.
Claims 1-2, 4-5, 9-11, 13-14, 18-20 are pending.
Response to Arguments
101 Arguments
Step 2A Prong 1
Applicant argues the Office has failed to identify any specific claim recitations as allegedly being directed to any specific sub-grouping, merely listing all the enumerated sub-groupings. Further, Applicant argues that the identified claim limitations do not recite a certain method of organizing human activities, as the claims do not recite fundamental economic practices or principles, commercial or legal interactions, or managing personal behavior or relationships or interactions between people.
This argument is not persuasive. The previous rejection (and the updated rejection herein) identifies the limitations that are directed towards the abstract idea. The identified limitations ultimately amount to provide personalized recommendations by generating customer behavior data based filtered transaction data and generating behavior mapping data to predict and provide personalized recommendations. This is considered to be advertising, marketing or sales activities or behaviors.
It is further noted that the amendments to the claims now specify that the training of the artificial neural network uses the historical state-action reward pairs (that are generated based on the customer behaviour data and a state-action-reward pair for each of a plurality of transactions and for each of one or more customers) and historically calculated customer behaviour to maximize rewards associated with each of the plurality of historical state-action pairs. This concept, the state action reward state action (SARSA), is an algorithm for learning a Markov decision process, a type of stochastic decision process. As such, the training of the artificial neural network model represents a mathematical concept that is interpreted to be part of the identified abstract idea, supra. The types of identified abstract ideas are considered together as a single abstract idea for analysis purposes.
Step 2A Prong 2
Applicant argues that the amended claims relate to a particular ANN-based recommendation model and recites particular non-generic training steps. By implementing an ANN-based recommendation model with optimally tuned hyperparameters selected from multiple neural network configurations, the system delivers higher accuracy for future state-action predictions, and removes the need for maintaining several neural network models with fewer hyperparameters, thereby reducing storage requirements and simplifying overall system complexity.
This argument is not persuasive. It is noted that the claims do not recite a methodology or a step of optimizing or tuning the hyperpameters. Instead, the optimal hyperparameter values are simply used when deploying the ANN-based recommendation model. In training the ANN, a plurality of neural network models are fetched together with corresponding parameter values. There is no indication that any of the parameter values of the fetched neural network models are optimized or tuned. Similarly, the claim does not specify any reduction in the number of hyperparameters, or even that the retrieved neural network models that are not optimized are removed or deleted; thus, it is not clear how the claim “removes the need for maintaining several neural network models with fewer hyperparameters, thereby reducing storage requirements and simplifying overall system complexity” as argued by Applicant.
Applicant argues that the amended claims are similar to Ex Parte Desjardins, submitting that the present claims reflect a similar improvement over previous systems as disclosed in paragraphs 5 and 33-36 of the specification.
This argument is not persuasive. Paragraph 5 identifies a shortcoming in the prior art, but the shortcoming is based on change in customer behaviour over time. In contrast, Ex Parte Desjardins
is eligible because of the manner in which the claimed machine learning model is trained, such that it effectively learns new tasks in succession whilst protecting knowledge about previous tasks (overcoming the problem of “catastrophic forgetting”). The specification does not address any issues with neural networks or training of neural networks that are comparable to Ex Parte Desjardins. Paragraph 33 only mention that an optimal model is selected upon training the neural network models, without detailing how the neural network models are trained. Paragraph 34 describes the output of the optimal model being a reward vector corresponding to each state-action pair from a trained neural network, and generating a corresponding matrix. The claims have been amended to remove any mention of a reward matrix, state matrix, and the reward vector is only recited as an output of the ANN-based recommendation model comprising a score value used when ranking a list of state-action pairs. The output of the ANN-based recommendation model does not constitute a technical solution per se, as the purpose of machine learning and neural network models is to output some sort of prediction or estimated score or value. Paragraph 35 describes that during model training, the state-action-reward behaviour database provides hyperparameters used to capture hidden patterns and unexpected future behaviour of customers. As noted previously, the claims do not identify or specify what hyperparameters are used or how they are tuned/optimized. They are simply “provided” during training of the neural network. The paragraphs cited by Applicant are insufficient to establish that the amended claims are similar to Ex Parte Desjardin in providing a technical solution to a technical problem with machine learning.
Step 2B
Applicant argues the claim recites inventive concepts that provide significantly more. Applicant argues that the claimed combination of elements in claim 1 meets the strong need in the field of relating to a Banking, Financial Services and Insurance (BFSI) company and/or retail sectors by utilizing a particular ANN-based recommendation model. Applicant further cites paragraph 90 of the specification, which states that the disclosed invention “try to overcome the technical problem of dynamically identifying change in customer behaviour and providing appropriate personalized recommendations”, as well as paragraph 3, identifying this problem affecting “companies in Banking, Financial Services, and Insurance (BFSI) and retail sectors aim at maximizing profits through data collection”, and paragraph 6, recognizing a “need in the present state of art for methods and systems to provide a context-based personalized recommendations to target customers based on real-time behaviour changes”.
This argument is not persuasive. The arguments presented by Applicant do not establish that the claimed invention solves or even addresses a technical problem. Rather, the problem is a business-related problem, in that prior art systems cannot identify change in customer behaviour to make personalized recommendations, the problem also being in banking, financial services and retail sectors seeking to maximize profits, and making personalized recommendations to target customers. As stated above and in the updated rejection, the identified abstract idea is most certainly belonging to the Certain Methods of Organizing Human Activity grouping, as the cited paragraphs of the specification indicate that the claimed invention evaluates commercial interactions including advertising, marketing or sales activities or behaviors, business relations; managing personal behavior.
In contrast, the claims do not recite any improvements to the functioning of a computer, technology or technical field (MPEP 2106.04(d)(1) and 2106.05(a)), apply the identified abstract idea with or by a particular machine (MPEP 2106.05(b)), effect a transformation or reduction of a particular article to a different state or thing (MPEP 2106.05(c)), add a specific limitation other than what is well-understood, routine or conventional in the field or adding unconventional steps that confine the claim to a particular useful application (MPEP 2106.05(d)), or apply or use the abstract idea in a meaningful way beyond generally linking the abstract idea to a particular technological environment (MPEP 2106.05(e)). Thus, the claims are not considered to provide or recite an inventive concept per MPEP 2106.05(I)(A) and 2106.05(II).
103 Arguments
Applicant presents several arguments on pages 27-35 with respect to amended claim 1 and the shortcomings of Vlassis, Nomula, Arora, Zhu, Ganduri.
These arguments are persuasive and the prior art rejection has been withdrawn.
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.
Claims 1-2, 4-5, 9-11, 13-14 and 18-20 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 – 2, 4 – 5, 9 (i.e. process), claim(s) 10 – 11, 13 – 14, 18 is/are drawn to system (i.e., a machine/manufacture – for the sake of this analysis these claims will be considered amended to be directed to one of the statutory categories), claim(s) 19 – 20 (i.e., program product). As such, claims 1-2, 4-5, 9-11, 13-14 and 18-20 are drawn to one of the statutory categories of invention.
Step 2A Prong 1: The claims recite an abstract idea of providing personalized recommendations by generating customer behavior data based filtered transaction data and generating behavior mapping data to predict and provide personalized recommendations , which is a certain method of organizing human activity (e.g. fundamental economic principles or practices including hedging, insurance, mitigating risk; commercial or legal interactions including agreements in the form of contracts, legal obligations, advertising, marketing or sales activities or behaviors, business relations; managing personal behavior or relationships or interactions between people including social activities, teaching, and following rules or instructions).
The claim limitations boldened below set forth the judicial exception:
Independent claims 1 and similarly 10 and 19:
generating customer behaviour data for each of a plurality of transactions and for each of one or more customers, based on time stamped customer transaction data and pre-defined product segmentation data;
generating a state-action-reward pair for each of the plurality of transactions and for each of one or more customers, based on the time stamped customer transaction data and the pre- defined product segmentation data wherein generating the state-action- reward pair includes processing the time stamped customer transaction data to extract a first state for each of the plurality of transactions and for each of the one or more customers and an action taken by each of one or more customers to perform a state transition from the first state to a subsequent state; and processing the pre-defined product segmentation data to determine reward associated with the state transition;
generating state-behaviour mapping data based on the customer behaviour data and the state-action-reward pair for each of the plurality of transactions and for each of one or more customers, wherein the state-behaviour mapping data comprises a plurality of historical state-action-reward pairs, a current state, and a current customer behaviour data for each of the one or more customers,
training an artificial neural network (ANN) based recommendation model based on the plurality of historical state-action-reward pairs and historically calculated customer behavior, wherein training comprises:
fetching a plurality of neural network models together with corresponding parameter values;
training the plurality of neural network models using the plurality of historical state-action-reward pairs and historically calculated customer behaviour, to maximize rewards associated with each of the plurality of historical state-action pairs, wherein the reward for a given state is determined based on a set of possible actions that transition a customer from the current state to a next state; and
deploying the ANN-based recommendation model with optimal hyper-parameter values from among the plurality of neural network models for predicting one or more future state-action pairs;
predicting a ranked list of next state-action pairs for a target customer, based on the current state of the target customer and the current customer behaviour data for the target customer, using the ANN based recommendation model, wherein predicting comprises
sending the current state and current customer behavior data from the state behaviour mapping data as input to the ANN-based recommendation model;
receiving a reward vector comprising a score value for each of a plurality of state-action pairs as output from the ANN-based recommendation model; and
ranking a list of next state-action pairs for the target customer based on the score value for each of a plurality of state-action pairs;
providing one or more personalized recommendations to the target customer based on the ranked list of next state-action pairs.
Step 2A Prong 2: The claim limitations recite the following additional elements that are beyond the judicial exception:
non-transitory computer-readable medium (claim 19)
a processor; (claim 10)
a memory communicatively coupled to the processor, wherein the memory stores processor instructions, which when executed by the processor, causes the processor to: (claim 10)
training an artificial neural network (ANN) based recommendation model based on the plurality of historical state-action pairs and historically calculated customer behaviour, wherein training comprises fetching a plurality of neural network models together with corresponding parameter values, training the plurality of neural network models using the plurality of historical state-action-reward pairs and historically calculated customer behaviour (Claim 1, 2 and 12)
deploying the ANN-based recommendation model with optimal hyper-parameter values from among the plurality of neural network models (Claim 1, 2 and 12)
input to the ANN-based recommendation model (Claim 1, 2 and 12)
output from the ANN-based recommendation model (Claim 1, 2 and 12)
using the ANN based recommendation model (Claim 1, 2 and 12)
None of the above cited additional elements are considered to provide an improvement to the functioning of a computer, technology or technical field (MPEP 2106.04(d)(1) and 2106.05(a)), effect a particular treatment or prophylaxis for a disease or medical condition (MPEP 2106.04(d)(2)), implement the identified abstract idea with a particular machine (MPEP 2106.05(b)), effect a transformation or reduction of a particular article to a different state or thing (MPEP 2106.05(c)), or apply or use the abstract idea in a meaningful way beyond generally linking the abstract idea to a particular technological environment (MPEP 2106.05(e)).
The following are recited at a high-level of generality (i.e., a generic server performing generic computer functions) such that it amounts no more than mere instructions to apply the exception using a generic computer component (see paragraphs 82-89 of the specification). Accordingly, these additional elements do not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea.
at least one or more processors
non-transitory computer-readable medium,
a memory communicatively coupled to the processor, wherein the memory stores processor instructions, which when executed by the processor, causes the processor to:
The following add the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea. See MPEP 2106.05(f).
ANN-based recommendation model
deploying the ANN-based recommendation model
predicting a ranked list of next state-action pairs for a target customer using the ANN-based recommendation model
using the ANN based recommendation model
The claim further recites “training an artificial neural network based recommendation model comprising fetching a plurality of neural network models together with corresponding parameter values and training the plurality of neural networks using the plurality of historical state-action-reward pairs and historically calculated customer behaviour and deploying the ANN-based recommendation model with optimal hyper-parameter values from among the plurality of neural network models for predicting one or more future state-action pairs”.
When given its broadest reasonable interpretation in light of the disclosure, the training of a machine learning model represents the application mathematical interrelationships between data, specifically using “state-action-reward pairs”. State action reward state action (SARSA) is an algorithm for learning a Markov decision process, a type of stochastic decision process. As such, the training of the artificial neural network model represents a mathematical concept that is interpreted to be part of the identified abstract idea, supra. The types of identified abstract ideas are considered together as a single abstract idea for analysis purposes.
The claim further recites the steps of using a trained neural network to “predict a ranked list of next state-action pairs for a target customer, comprising sending the current state and current customer behaviour from the state behaviour mapping data as input to the ANN-based recommendation model, receiving a reward vector comprising a score value for each of a plurality of state-action pairs as output from the ANN-based recommendation model and ranking a list of next state-action pairs for the target customer based on the score value for each of a plurality of state-action pairs”.
This represents mere instructions to implement the abstract idea on a generic computer. Implementing an abstract idea using a generic computer or components thereof does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. See, e.g., Recentive Analytics, Inc. v. Fox Corp., No. 2023-2437 at 10 (Fed. Cir. April 18, 2025) (finding that claims that do no more than apply established methods of machine learning to a new data environment are ineligible). Alternatively, or in addition, the implementation of the trained machine learning model to predict a ranked list of next state-action pairs merely confines the use of the abstract idea (i.e., the trained neural network model) to a particular technological environment or field of use (to provide personalized recommendations based on customer transaction data and pre-defined product segmentation data) and thus fails to add an inventive concept to the claims.
Further, MPEP 2106.04(d)(I) and MPEP 2106.05(A) indicate that merely “generally linking” the abstract idea to a particular technological environment or field of use cannot provide a practical application or significantly more.
The claim further recites the additional elements of providing one or more personalized recommendations to the target customer based on the ranked list of next state-action pairs. This transmitting/providing step is recited at a high level of generality (i.e., as a general means of transmitting/providing data) and amounts to the mere transmission of data, which is a form of extra-solution activity. MPEP 2106.04(d)(I) indicates that extra-solution data gathering activity cannot provide a practical application. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application.
Step 2B
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the use of generic computer hardware components or artificial neural networks (even after being trained) to perform the noted steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept (“significantly more”).
As discussed above with respect to integration of the abstract idea into a practical application, the additional element of using the trained artificial neural network model to predict a ranked list of next state-action pairs was found to represent mere instructions to implement the abstract idea on a generic computer and/or confine the use of the abstract idea (i.e., the trained model) to a particular technological environment or field of use. This has been re-evaluated under the “significantly more” analysis and determined to be insufficient to provide significantly more. MPEP 2106.05(I) indicates that mere instructions to implement the abstract idea on a generic computer and/or confining the use of the abstract idea to a particular technological environment or field of use cannot provide significantly more. See also Recentive Analytics, Inc. v. Fox Corp., No. 2023-2437 at 17 (Fed. Cir. April 18, 2025) (finding that applying machine learning to an abstract idea does not transform a claim into something significantly more).
Also, as discussed above with respect to integration of the abstract idea into a practical application, the additional elements of providing the one or more personalized recommendations to the target customer were considered extra-solution activity. This has been re-evaluated under the “significantly more” analysis and determined to be well-understood, routine, conventional activity in the field. MPEP 2016.05(d)(II) indicates that receiving and/or transmitting data over a network has been held by the courts to be well-understood, routine, conventional activity (citing Symantec, TLI Communications, OIP Techs., and buySAFE). Well-understood, routine, conventional activity cannot provide an inventive concept (“significantly more”). As such the claim is not patent eligible.
As discussed with respect to step 2A prong 2 above, the additional elements of non-transitory computer-readable medium, a processor; a memory communicatively coupled to the processor, wherein the memory stores processor instructions, which when executed by the processor, causes the processor to, “training an artificial neural network (ANN) based recommendation model based on the state-behaviour mapping data, wherein training comprises fetching a plurality of neural network models together with corresponding parameter values, training the plurality of neural network models using the plurality of historical state-action-reward pairs and historically calculated customer behaviour, deploying the ANN-based recommendation model with optimal hyper-parameters from among the plurality of neural network models for predicting one or more future state-action pairs” and “predicting a ranked list of next state-action pairs for a target customer, based on the current state of the target customer and the current customer behaviour data for the target customer using the ANN based recommendation model wherein predicting comprises sending the current state and current customer behaviour from the state behaviour mapping data as input to the ANN-based recommendation model, receiving a reward vector comprising a score value for each of a plurality of state-action pairs as output from the ANN-based recommendation model and ranking a list of next state-action pairs for the target customer based on the score value for each of a plurality of state-action pairs” are mere instructions to apply an exception, and do not integrate a judicial exception into a practical application at step 2A or provide an inventive concept at step 2B.
According to MPEP 2106 a conclusion that an additional element is mere instructions to apply an exception under step 2A should be re-evaluated at step 2B. Thus, the additional elements identified above are re-evaluated to determine whether they constitute significantly more. Examiner finds that the additional elements are simply the use of a computer in its ordinary capacity and does not provide significantly more. See Affinity Labs v. DirecTV, 838 F.3d 1253, 1262 and MPEP 2106.05(f). For example, the additional elements only provide a result-oriented solution and lack details as to how the computer performs the abstract idea above, which is equivalent to “apply it”. See Alice Corp. v. CLS Bank, 134 S. Ct. 2347, 2357 and MPEP 2106.05(f).
Therefore, when considering all the additional claim elements both individually and as an ordered combination, Examiner finds that the claim does not amount to significantly more than the exception.
Dependent Claims 2, 4-5, 9, 11, 13-14 and 18, and 20 further narrow the abstract idea and/or the additional elements disclosed in the claims have been addressed above.
Subject Matter Free of the Prior Art
Independent claims 1, 10, and 19 recite subject matter that is free of the prior art. These claims recite the combination of the following limitations:
generating customer behaviour data for each of a plurality of transactions and for each of one or more customers, based on time stamped customer transaction data and pre-defined product segmentation data;
generating a state-action-reward pair for each of the plurality of transactions and for each of one or more customers, based on the time stamped customer transaction data and the pre-defined product segmentation data, wherein generating the state-action-reward pair includes:
processing the time stamped customer transaction data to extract a first state for each of the plurality of transactions and for each of the one or more customers and an action taken by each of one or more customers to perform a state transition from the first state to a subsequent state; and
processing the pre-defined product segmentation data to determine a reward associated with the state transition;
generating, by the recommendation system, state-behaviour mapping data based on the customer behaviour data and the state-action-reward pair for each of the plurality of transactions and for each of one or more customers,
wherein the state-behaviour mapping data comprises a plurality of historical state-action-reward pairs, a current state, and a current customer behaviour data for each of the one or more customers
training an artificial neural network based recommendation mode model based on the plurality of historical state-action-reward pairs and historically calculated customer behaviour, wherein training comprises:
fetchinq a plurality of neural network models together with correspondinq parameter values;
training the plurality of neural network models using the plurality of historical state-action-reward pairs and historically calculated customer behaviour, to maximize rewards associated with each of the plurality of historical state-action pairs,
wherein the reward for a given state is determined based on a set of possible actions that transition a customer from the current state to a next state; and
deployinq the ANN-based recommendation model with optimal hyper- parameter values from amonq the plurality of neural network models for predictinq one or more future state-action pairs;
predicting one or more future state-action pairs by maximizing reward for each of the plurality of historical state-action pairs; predicting, by the recommendation system, a ranked list of next state-action pairs for a target customer, based on the current state of the target customer and the current customer behaviour data for the target customer, using the ANN-based recommendation model, wherein predictinq comprises:
sending the current state and current customer behaviour from the state behaviour mappinq data as input to the ANN-based recommendation model;
receivinq a reward vector comprisinq a score value for each of a plurality of state-action pairs as output from the ANN-based recommendation model; and
rankinq a list of next state-action pairs for the target customer based on the score value for each of a plurality of state-action pairs; and
providing, by the recommendation system, one or more personalized recommendations to the target customer based on the ranked list of next state-action pairs
Vlassis (US PG Pub 20180165590) teaches applying machine learning techniques to generate personalized recommendations based on a transition matrix of a user describing the propensity to transition from a current state to a different state. The transition matrix serves as the set of conditional transition probabilities between states, and may incorporate a reward function to be optimized in the user model, where the reward function specifies a desirability for the user to transition to particular states. The recommendation system utilizes current and historical user behavior data. The transition matrix (also referred to as a Markov chain) reflects the historical probability of user transition from one of a collection of states to another state.
Nomula (US PG Pub 20190139092) teaches utilizing an iterative learning process for a neural network, as well as a ranking module to rank results according to their scores for recommendation to the user. Previous transaction data of the user and item segment information may be used in the training of the neural network. A user profile vector is generated to characterize the user’s behavior, and can also be used to generate content and advertisement recommendations using Deep Reinforcement Learning.
Zhu (US PG Pub 20180165745) teaches reinforcement learning processing using a Markov Decision Making process and calculates a reward value corresponding to a state-action pair.
Sequeira (US PG Pub 20200320435) teaches a reinforcement learning process that maps actions from a given state and from states to actions using state-action pair value functions, while maximizing a reward received after executing some action leading to choosing one action over others.
Though the cited prior art references teach portions of the invention set forth in the independent claims, no prior art reference, separately or in combination, teach or render obvious the combination of the limitations recited in the independent claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PETER H CHOI whose telephone number is (469)295-9171. The examiner can normally be reached M-Th 9am-7pm.
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) at http://www.uspto.gov/interviewpractice.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/PETER H CHOI/ Supervisory Patent Examiner, Art Unit 3681
Prosecution Insights