DETAILED ACTION
Status of the Claims
The following is a Non-final Office Action in response to amendments and remarks filed 04 March 2026.
Claims 1, 10, and 17 have been amended.
Claims 1-3, 5-6, and 8-22 are pending have been examined.
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 08 April 2026 has been entered.
Response to Arguments
Applicants argue that the 35 U.S.C. 101 rejection under the Alice Corp. vs. CLS Bank Int’l be withdrawn; however the Examiner respectfully disagrees. The Examiner notes that in order to be patent eligible under 35 U.S.C. 101, the claims must be directed towards a patent eligible concept, which, the instant claims are not directed. Contrary to Applicants’ assertion that the claims are not a certain method of organizing human activity, the Examiner notes that generating contestation documents based upon generated machine learning scores that indicate the likelihood of successfully winning a chargeback request is an abstract idea of certain methods of organizing human activities (i.e. commercial or legal interactions (including agreements in the form of contracts; legal obligations; advertising, marketing or sales activities or behaviors; business relations)) as well as the abstract idea of performing computations in accordance with a mathematical formula on that data (generating a score) is a function that financial institutions, merchants, insurance companies, fraud investigators etc. have traditionally performed/provided for consumers. The Examiner again notes that the claims are drawn to organizing human activities, while some may be based upon a mathematical concept. The Examiner notes that the generating contestation documents based upon generated machine learning scores that indicate the likelihood of successfully winning a chargeback request which is an abstract idea of organizing human activities. The claims quite literally organize commercial interactions/sales activities//business relations (transactions). The entirety of the specification is discussing how to capture this transaction data over time periods which is, again, quite literally commercial interactions such as sales activities or behavior; business relation while some of the limitations may involve a mathematical concept being performed on the organized human activities (i.e. a generated score). Next, the claims are not directed to a practical application of the concept. The claims do not result in improvements to the functioning of a computer or to any other technology or technical field. They do not effect a particular treatment for a disease. They are not applied with or by a particular machine. They do not effect a transformation or reduction of a particular article to a different state or thing. And they are not applied in some other meaningful way beyond generally linking the use of the judicial exception (i.e., generating contestation documents based upon generated machine learning scores that indicate the likelihood of successfully winning a chargeback request) to a particular technological environment (i.e., with the use of a generic computer or computing components and machine learning). Here, again as noted in the previous rejection, mere instructions to apply an exception using a generic computer component cannot provide an inventive concept - MPEP 2016.05(f). The recitation of “using at least one machine learning model” in the limitations also merely indicates a field of use or technological environment in which the judicial exception is performed. Although the additional element “using at least one machine learning model” limits the identified judicial exceptions, this type of limitation merely confines the use of the abstract idea to a particular technological environment (machine learning) and thus fails to add an inventive concept to the claims. See MPEP 2106.05(h). Accordingly, the combination of these additional elements does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim(s) is/are not patent eligible. As such, this argument is not persuasive, and the rejection not overcome.
Applicant’s next argue that the claims are eligible as the claims cannot be “practically performed in the human mind;”(due to the recitation of machine learning and vectors) however the Examiner respectfully disagrees. As an intimal note, the general theory behind machine learning is to model a computer around the human brain, such that the computer learns as our brains do. Secondly, while the specification may discuss sophisticated techniques and advanced functions such as machine learning, it is the claims that are deemed eligible or ineligible under §101. Here, the claims are more of a generalized guideline of how to arrange a software model to implement the overarching abstract idea. The claim(s) is/are not patent eligible and the rejection not withdrawn.
Applicants again argue that the claims are similar to the Enfish decision; however the Examiner respectfully disagrees. Unlike Enfish, the instant claims are clearly focused on the combination of those abstract-idea processes. Here, by contrast, the claims are not directed to an improvement in the way computers operate. The advance the claims purport to make is a process of generating contestation documents based upon generated machine learning scores that indicate the likelihood of successfully winning a chargeback request i.e. organizing human activities such as transaction data for analysis and generation of documentation thereof, and not any particular assertedly inventive technology for performing those functions. They are therefore still directed to an abstract idea. Again, unlike Enfish, the instant claims focus not on such an improvement to the computers as tools, but on certainly independent abstract ideas that use the computers as tools and therefore do not amount to significantly more than the abstract ideas themselves. Merely requiring the selection and manipulation of information—to provide a “humanly comprehensible” amount of information useful for users, by itself does not transform the otherwise-abstract processes of information collection and analysis. Instead, the claims are more analogous to those in FairWarningIP, LLC v. Iatric Sys., Inc., 839 F.3d 1089 (Fed. Cir. 2016), wherein claims reciting “a few possible rules to analyze audit log data” were found directed to an abstract idea because they asked “the same questions (though perhaps phrased with different words) that humans in analogous situations detecting fraud have asked for decades.” FairWarning, 839 F.3d at 1094, 1095. As such, this argument is not persuasive, the rejection not withdrawn.
Applicant next argues the claims are integrated into a practical application due to being more efficient; however, again, this argument appears to be whether or not the use of computer or computing components for increased speed and efficiency via machine learning results in an improvement; however the Examiner respectfully disagrees. Nor, in addressing the second step of Alice, does claiming the improved speed or efficiency inherent with applying the abstract idea on a computer provide a sufficient inventive concept. See Bancorp Servs., LLC v. Sun Life Assurance Co. of Can., 687 F.3d 1266, 1278 (Fed. Cir. 2012) (“[T]he fact that the required calculations could be performed more efficiently via a computer does not materially alter the patent eligibility of the claimed subject matter.”); CLS Bank, Int’l v. Alice Corp., 717 F.3d 1269, 1286 (Fed. Cir. 2013) (en banc) aff’d, 134 S. Ct. 2347 (2014) (“[S]imply appending generic computer functionality to lend speed or efficiency to the performance of an otherwise abstract concept does not meaningfully limit claim scope for purposes of patent eligibility.” (citations omitted)). As such, this argument is not persuasive and the rejection not withdrawn.
Applicant argues that the cited references do not teach or suggest the newly amended claims with respect to filtering by time periods; however the Examiner respectfully disagrees. While not previously cited (but noted in the Advisory Action 11 March 2026), Ramanjani does in fact disclose the ability to analyze transactions over pre-defined or specific time periods and (Ramanjani ¶62-¶63 and ¶84) as noted in the updated rejection below. As such this argument is not persuasive and the rejection not overcome.
In response to arguments in reference to any depending claims that have not been individually addressed, all rejections made towards these dependent claims are maintained due to a lack of reply by the Applicants in regards to distinctly and specifically pointing out the supposed errors in the Examiner's prior office action (37 CFR 1.111). The Examiner asserts that the Applicants only argue that the dependent claims should be allowable because the independent claims are unobvious and patentable over the prior art..
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 1-3, 5-6, and 8-22 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims are directed to a process (an act, or series of acts or steps), a machine (a concrete thing, consisting of parts, or of certain devices and combination of devices), and a manufacture (an article produced from raw or prepared materials by giving these materials new forms, qualities, properties, or combinations, whether by hand labor or by machinery). Thus, each of the claims falls within one of the four statutory categories (Step 1). However, the claim(s) recite(s) generating contestation documents based upon generated machine learning scores that indicate the likelihood of successfully winning a chargeback request which is an abstract idea of organizing human activities and/or a mental process on gathered transactional data.
The limitations of “generating, at least one machine learning score, wherein the at least one machine learning score indicates a likelihood of successfully winning the chargeback request wherein generating the at least one machine learning score further comprises: filtering the transaction data and the service provider data according to date parameters to generate filter data, wherein data within a first predefined time period is excluded and data after the first predefined time period is aggregated over one or more second predefined time periods; generating an input vector the filtered data and historical data, where the historical data comprises data collected over a third predefined time period, and wherein a size of the input vector is reduced relative to a size of the filtered transaction data; and generating, by passing the input vector through an input layer, a plurality of hidden layers, and an output layer of the at least one machine learning model, the at least one machine learning score; determining, using the at least one machine learning score, the transaction data, the service provider data, and at least one dispute processing rule in a plurality of dispute processing rules the likelihood of successfully winning the chargeback request; based on the likelihood of winning the chargeback request, generating from the at least one template a contestation document; and submitting the contestation document to contest the chargeback request,” as drafted, is a process that, under its broadest reasonable interpretation, organizing human activities--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) or a mental process—concepts performed in the human mind (including an observation, evaluation, judgment, opinion) but for the recitation of generic computer components (Step 2A Prong 1). That is, other than reciting “using at least one machine learning model in a machine learning framework executing on the processor,” (or “A system comprising: a non-transitory memory storing instructions; and one or more hardware processors coupled to the non-transitory memory and configured to read the instructions from the non-transitory memory to cause the system to perform operations comprising:” in claim 10 or “A non-transitory machine-readable medium having stored thereon machine- readable instructions executable to cause a machine to perform operations comprising:” of claim 17) nothing in the claim element precludes the step from the organizing human activities or mental process grouping. For example, but for the “using at least one machine learning model in a machine learning framework executing on the processor,” (or “A system comprising: a non-transitory memory storing instructions; and one or more hardware processors coupled to the non-transitory memory and configured to read the instructions from the non-transitory memory to cause the system to perform operations comprising:” in claim 10 or “A non-transitory machine-readable medium having stored thereon machine- readable instructions executable to cause a machine to perform operations comprising:” of claim 17)” language, “generating,” “filtering,” “generating,” generating,” and “determining” in the context of this claim encompasses the user manually scoring and determining likelihood of winning a chargeback request based upon the obtained transactional data which is mental process (such as statistical analysis) generate a score for which a decision is made and templates be generated and submitted (organizing human activities such as legal or business documentation generated based upon transactions and the score). If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation as one of the methods of organizing human activities or mental processes, while some of the limitations may be organizing human activities, but for the recitation of generic computer components, then it falls within the “Certain Methods of Organizing Human Activities” and/or “Mental Process” grouping of abstract ideas. Accordingly, the claim(s) recite(s) an abstract idea (Step 2A, Prong One: YES). However, if possible, the Examiner should consider the limitations together as a single abstract idea rather than as a plurality of separate abstract ideas to be analyzed individually. “For example, in a claim that includes a series of steps that recite mental steps as well as a mathematical calculation, an Examiner should identify the claim as reciting both a mental process and a mathematical concept for Step 2A, Prong One to make the analysis clear on the record.” MPEP 2106.04, subsection II.B. Under such circumstances, however, the Supreme Court has treated such claims in the same manner as claims reciting a single judicial exception. Id. (discussing Bilski v. Kappos, 561 U.S. 593 (2010)). Here, the limitations are considered together as a single abstract idea for further analysis.
This judicial exception is not integrated into a practical application (Step 2A Prong Two). The “receiving, at a dispute processing system executing on a processor, a chargeback request for a transaction; receiving transaction data and service provider data associated with the transaction; incorporating the transaction data and the service provider data into an at least one template” are simply insignificant extrasolution data gathering activities. Next, the claim only recites one additional element – using at least one machine learning model in a machine learning framework executing on the processor or A system comprising: a non-transitory memory storing instructions; and one or more hardware processors coupled to the non-transitory memory and configured to read the instructions from the non-transitory memory to cause the system to perform operations or a machine to perform the steps. The processor, system, and machine in the steps is recited at a high-level of generality (i.e., as a generic processor performing a generic computer function of scoring data and filling forms) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Specifically the claims amount to nothing more than an instruction to apply the abstract idea using a generic computer or invoking computers as tools by adding 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 2106.04(d)(I) discussing MPEP 2106.05(f). The recitation of “using at least one machine learning model” in the limitations also merely indicates a field of use or technological environment in which the judicial exception is performed. Although the additional element “using at least one machine learning model” limits the identified judicial exceptions, this type of limitation merely confines the use of the abstract idea to a particular technological environment (machine learning) and thus fails to add an inventive concept to the claims. See MPEP 2106.05(h). Accordingly, the combination of these additional elements does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea, even when considered as a whole (Step 2A Prong Two: NO).
The claim does not include a combination of additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B). As discussed above with respect to integration of the abstract idea into a practical application (Step 2A Prong 2), the combination of additional elements of using a processor, system, and machine to perform the 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. Reevaluating here in step 2B, the “receiving, at a dispute processing system executing on a processor, a chargeback request for a transaction; receiving transaction data and service provider data associated with the transaction; incorporating the transaction data and the service provider data into an at least one template” step(s) which are insignificant extrasolution activities are also determined to be well-understood, routine and conventional activity in the field. The Symantec, TLI, and OIP Techs court decisions in MPEP 2106.05(d)(II) indicate that the mere receipt or transmission of data over a network is well-understood, routine, and conventional function when it is claimed in a merely generic manner (as is here). Therefore, when considering the additional elements alone, and in combination, there is no inventive concept in the claim. As such, the claim(s) is/are not patent eligible, even when considered as a whole (Step 2B: NO).
Claims 2, 5-6, 8, 12-15, and 18-22 recite(s) the additional limitation(s) further limiting the score, which is still directed towards the abstract idea previously identified and is not an inventive concept that meaningfully limits the abstract idea. Again, as discussed with respect to claims 1, 10, and 17, the claims are simply limitations which are no more than mere instructions to apply the exception using a computer or with computing components. Accordingly, the additional element(s) does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. Even when considered as a whole, the claims do not integrate the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Claims 3, 9, 11, and 16 recite(s) the additional limitation(s) further including some sort of post solution data output/update which is not an inventive concept that meaningfully limits the abstract idea. Again, as discussed with respect to claims 1, 10, and 17, the claims are simply limitations which are no more than mere instructions to apply the exception using a computer or with computing components. Accordingly, the additional element(s) does not integrate the abstract idea into a practical application because it does not impose any meaningful limits on practicing the abstract idea. Even when considered as a whole, the claims do not integrate the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Claims 1-3, 5-6, and 8-22 are therefore not eligible subject matter, even when considered as a whole.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-3, 5-6, 9-12, 16-19, and 21-22 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ramanjani et al. (US PG Pub. 2023/0281635) and further in view of Ross et al. (US PG Pub. 2022/0309507).
As per claims 1, 10, and 17, Ramanjani discloses a method, system comprising: a non-transitory memory storing instructions; and one or more hardware processors coupled to the non-transitory memory and configured to read the instructions from the non-transitory memory to cause the system to perform operations comprising; non-transitory machine-readable medium having stored thereon machine- readable instructions executable to cause a machine to perform operations comprising (apparatus, at least one processor, non-transitory computer readable storage medium, interface of at least one user device, Ramanjani ¶7-¶9; system, ¶33):
receiving, at a dispute processing system executing on a processor, a chargeback request for a transaction (receive data associated with at least one disputed transaction for at least one user, wherein the received data includes user-specific information, merchant-specific information, or a combination thereof; process the received data to calculate a probability of success in a chargeback representment for the at least one disputed transaction, Ramanjani ¶8; The cardholder may initiate a transaction dispute, e.g., chargeback, with the issuing bank to return some or all of the funds associated with the disputed transaction to the account corresponding to the payment card. Chargebacks can also occur as a result of friendly fraud, where the transaction was authorized by the cardholder but the cardholder later attempts to fraudulently (i.e., either mistakenly or maliciously) reverse the charges. Such types of chargebacks can be financially detrimental to the merchant and may affect the ability of the merchants to accept certain payment vehicles in the future, ¶31) (Examiner notes the transaction dispute initiation as the equivalent to the chargeback request);
receiving transaction data and service provider data associated with the transaction (receive data associated with at least one disputed transaction for at least one user, wherein the received data includes user-specific information, merchant-specific information, or a combination thereof; process the received data to calculate a probability of success in a chargeback representment for the at least one disputed transaction, Ramanjani ¶8);
incorporating the transaction data and the service provider data into an at least one template (In one embodiment, data collection module 201 may automatically collect relevant data of users, e.g., customers, merchants, etc., associated with the disputed transaction through various data collection techniques. For example, data collection module 201 may use a web-crawling component to access various databases or other information sources to determine the online activities information of the users. Data collection module 201 may be programmed to collect, in real-time, historical information, contextual information, or a combination thereof pertaining to the users, so that identification, analysis, response, monitoring, and control may be performed using the most recent data. In one example embodiment, data collection module 201 may include various software applications, e.g., data mining applications in Extended Meta Language (XML), that automatically search for and return relevant information pertaining to the users. In one embodiment, data collection module 201 may parse and arrange the data into a common format that can be easily processed by other modules and platforms. In one embodiment, data collection module 201 may collect, in real-time, various evidence to support the chargeback dispute from various information sources and/or database 111. The collected evidence may include any type of documentary evidence, e.g., proof of delivery, proof of shipment, or other compelling evidence associated with the cardholder, merchant, or the disputed payment vehicle transaction, Ramanjani ¶48) (Examiner notes the arrangement of data into a common format as the ability to incorporate the transaction/service provider data into a template);
generating, using at least one machine learning model in a machine learning framework executing on the processor, at least one machine learning score, wherein the at least one machine learning score indicates a likelihood of successfully winning the chargeback request (the calculated win probability scores based on machine learning predictions with represent/don't represent business rules to generate recommendations on chargeback representment. For example, data analytics platform 109 may compare the calculated win probability of 51% to a minimum threshold probability requirement, e.g., 72%. Data analytics platform 109 may not recommend chargeback representment because the calculated probability of 51% is below the minimum threshold requirement. Data analytics platform 109 may recommend chargeback representment if the win probability is equal to or higher than 72%, Ramanjani ¶54; activate a plurality of machine learning algorithms, ¶71);
filtering the transaction data and the service provider data according to date parameters to generate filter data, wherein data within a first predefined time period is excluded and data after the first predefined time period is aggregated over one or more second predefined time periods (In calculating the probability of re-occurrence, probability module 305 may generate a matrix that represents a pattern of re-occurrence based, at least in part, on historical data, behavior data, or a combination thereof. In one example embodiment, probability module 305 may determine that dishonest customers repeat their fraudulent behavior within a particular time period, e.g., 20 days, based on historical data. Probability module 305 may assign a score or a weighting value to historical transaction data, device information, account information, or a combination thereof associated with the dishonest customer. In another example embodiment, probability module 305 may determine a higher occurrence of fraudulent transactions during a specific time period, e.g., day, month, or time, based on historical data. Probability module 305 may assign a score or a weighting value to temporal information, historical transaction data, device information, account information, or a combination thereof associated with the disputed transaction. Probability module 305 may then implement the assigned scores and weighting value to generate a probability result, Ramanjani ¶62-¶64; In step 437, data analysis platform 109 may determine a re-occurrence of at least one disputed transaction within a pre-determined time period based, at least in part, on the received data, behavior data, or a combination thereof associated with at least one user. In one embodiment, data analysis platform 109 may determine frequency information associated with a plurality of data transactions, and convert the frequency information into a frequency variable. In one example embodiment, data analysis platform 109 may dynamically track the amount of the transactions, the time at which the transactions occurred, the categories of the transactions, frequency of the transactions, most frequently visited ATMs, or the like for an individual account. Any deviation from such behaviors indicates a higher probability of fraud. Data analysis platform 109 may determine a re-occurrence of a disputed transaction in the future based on the frequency information. In some embodiments, the frequency variable may also be used with at least one other variable to predict fraudulent activity, ¶84; dynamic based upon time period, ¶89)
determining, using the at least one machine learning score, the transaction data, the service provider data, and at least one dispute processing rule in a plurality of dispute processing rules the likelihood of successfully winning the chargeback request (In one embodiment, data analysis platform 109 may receive data associated with at least one disputed transaction for at least one user, wherein the data includes chargeback data. Data analysis platform 109 may process the data to calculate a probability of success in a chargeback representment for the at least one disputed transaction. Data analysis platform 109 may then calculate a prediction based, at least in part, on the probability of success, one or more results from a plurality of machine learning models, or a combination thereof. Data analysis platform 109 may generate a presentation of at least one recommendation on the chargeback representment based, at least in part, on the prediction in a user interface of at least one device associated with the at least one user, Ramanjani ¶40);
based on the likelihood of winning the chargeback request, generating from the at least one template a contestation document (In step 503, data analysis platform 109 may process the chargeback data via machine learning mechanisms. In one example embodiment, data analysis platform 109 may analyze the chargeback data based, at least in part on, historical chargeback data and economic feasibility data to determine whether to automatically initiate a chargeback representment, automatically queue or flag the chargeback for further review, or automatically accept financial liability for the chargeback. In one embodiment, data analysis platform 109 may calculate the win/loss probability and prediction for a chargeback representment, Ramanjani ¶87) (Examiner interprets the automatic initiation of the chargeback representment as the ability to generate a contestation document from the at least one template); and
submitting the contestation document to contest the chargeback request (automatically initiate a chargeback representment, Ramanjani ¶87).
Ramanjani does not expressly disclose wherein generating the at least one machine learning score further comprises: generating an input vector the filtered data and historical data, where the historical data comprises data collected over a third predefined time period, and wherein a size of the input vector is reduced relative to a size of the filtered transaction data; and generating, by passing the input vector through an input layer, a plurality of hidden layers, and an output layer of the at least one machine learning model, the at least one machine learning score.
However, Thomas teaches wherein generating the at least one machine learning score further comprises: generating an input vector the filtered data and historical data, where the historical data comprises data collected over a third predefined time period, and wherein a size of the input vector is reduced relative to a size of the filtered transaction data; and generating, by passing the input vector through an input layer, a plurality of hidden layers, and an output layer of the at least one machine learning model, the at least one machine learning score (Thus, the ML engine may provide output classifications of evidence for use in and response to the dispute based on the input dispute attributes. The dispute resolution platform may display the recommended evidence via one or more user interfaces (UIs), which may present both the probabilities and/or tiers associated with the evidence and the evidence's success rates, Ross ¶26; input layers, hidden layers, ¶41-¶43; Thus, the training data may be used as input/output data sets that allow for ML models 134 to make classifications based on input attributes. The output classifications for an ML model trained for evidence classifications 136 may be classifications of the categories and/or codes for uncategorized evidence. The output classifications for an ML model trained for evidence recommendations 138 may be classifications of the categories and/or types of evidence recommended for a dispute or dispute type (e.g., based on dispute details or attributes) from win probabilities of evidence classifications 136. Evidence classifications 136 may include evidence previously classified into categories, such as by a data scientist, administrator, machine or other entity that classifies the evidence to categories or otherwise annotates the training data for evidence classification. Evidence classifications 136 may also include the uncategorized data now classified into categories by one or more of ML models 134. Thereafter, evidence classifications 136 may be used to further train one or more of ML models 134, which the provides evidence recommendations 138 based on received disputes, ¶44; At step 406, a first ML model for evidence classification is trained using the extracted text data. The first ML model may be trained using attributes from the evidence classification training data, including those attributes identified from the extracted text data (e.g., words, phrases, and the like that may be converted to vectors for vector comparisons and the like when training layers of a ML model). Further, the input attributes may include the evidence categories corresponding to the extracted text data. For example, the evidence in the training data may be classified as categories or types corresponding to codes or identifiers such that the extracted text data may then be correlated to these categories. This data therefore allows training of the first ML model to make classifications of text from other evidence into categories from the training data, ¶72) (Examiner notes the size of the input vector is relative due to the size of the filtered data as taught by Ramanjani).
Both the Ross and Ramanjani references are analogous in that both are directed towards/concerned with dispute resolution. Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to use Thomas’ ability to have layers in machine learning algorithms in Ramanjani’s system to improve the system and method with reasonable expectation that this would result in a dispute resolution management system that is able to process more evidence for merchants.
The motivation being that service providers, such as online transaction processors, may provide electronic transaction processing services to merchants and customers. As part of the electronic transaction processing services, a service provider may also provide dispute resolution services, such as when an error or disagreement occurs during transaction processing. These disputes may be based on exchanged values, item quality or quantity, delivery, and the like. However, in order to resolve these disputes, the service provider's dispute resolution service may require evidence to support the merchant's and/or customer's position regarding the dispute and determine how to resolve the dispute in favor of one of the sides. Without knowing what evidence causes a positive resolution of the dispute on behalf of the merchant, the merchant may not locate and upload the proper evidence. Further, current systems do not recommend the available types of evidence and/or provide a data storage of available evidence for merchants to easily select and upload evidence during a dispute. This can cause unnecessary time and computing resources for both the service provider and the merchant from notifying the merchant that the evidence is not proper, the merchant determining and locating other evidence, resubmitting new evidence, and processing the resubmission of new evidence (Ross ¶2).
Ramanjani further discloses the ability to utilize historical data, where the historical data comprises data collected over a predefined time period (In one embodiment, probability module 305 may implement machine-learned rules to calculate the likelihood a user may win a chargeback representment. The calculation may include determining a win probability based, at least in part, on an analysis of data of the actual transaction and historical big data, e.g., data of past transactions. In one embodiment, historical data may include customer-specific information indicative of how often a specific customer successfully or unsuccessfully disputed online transactions. In another embodiment, historical data may include merchant-specific information indicative of how often a specific merchant successfully disputed particular types of chargebacks. In a further embodiment, historical data may include global historical win rates (or percentages) indicative of how often, based on the particular chargeback reason code, a group of merchants, e.g., a global population of merchants, successfully disputed particular types of chargebacks, Ramanjani ¶60) (Examiner interprets the ability to include some sort of history as the ability to include data collected over a predetermined time period).
As per claim 2, Ramanjani and Ross disclose as shown above with respect to claim 1. Ramanjani further discloses determining, using the at least one machine learning score, the transaction data, the service provider data, and the at least one dispute processing rule indicating a likelihood of successfully winning a pre-arbitration process (whether a chargeback reached the presentment stage, the re-presentment stage, or the arbitration stage, and transaction date and time, Ramanjani ¶86; In step 503, data analysis platform 109 may process the chargeback data via machine learning mechanisms. In one example embodiment, data analysis platform 109 may analyze the chargeback data based, at least in part on, historical chargeback data and economic feasibility data to determine whether to automatically initiate a chargeback representment, automatically queue or flag the chargeback for further review, or automatically accept financial liability for the chargeback. In one embodiment, data analysis platform 109 may calculate the win/loss probability and prediction for a chargeback representment, ¶87) (Examiner interprets the ability to score each stage as including the likelihood of successfully winning a pre-arbitration process).
As per claims 3 and 11-12, Ramanjani and Ross disclose as shown above with respect to claims 1 and 10. Ramanjani further discloses transmitting, to a dispute processing system interface, a message indicating a status of the chargeback request; and displaying, on the dispute processing system interface, the status of the chargeback request and details of the chargeback request; transmitting, to a dispute processing system interface, a plurality of messages, each message indicating an updated status of the chargeback request; and updating status of the chargeback request displayed on the dispute processing system interface upon receipt of the each message (In one embodiment, recommendation module 307 may generate a recommended action in response to the disputed transaction, e.g., chargeback based on the data received from machine learning module 207 and prediction module 209. In another embodiment, recommendation module 307 may collaborate with user interface module 211 to generate a representment notification to inform the users of the decision to dispute the chargeback or accept financial liability. In one example embodiment, recommendation module 307 may determine that merchants have a higher probability of winning chargeback representments if the dispute is filed within a pre-defined time period, e.g., 14 days, based on historical data. Recommendation module 307 may recommend a merchant to dispute the chargeback within the pre-defined time period, Ramanjani ¶63).
As per claim 5, Ramanjani and Ross disclose as shown above with respect to claim 1. Ross further teaches generating, using the input vector, plurality of input layer outputs; applying, at each neuron in a hidden layer of the plurality of hidden layers, a weighted matrix and an activation function to each input layer output, to generate a hidden layer output; and generating, at the output layer, the at least one machine learning score from the hidden layer output (input layers, hidden layers, Ross ¶41-¶43; Thus, the training data may be used as input/output data sets that allow for ML models 134 to make classifications based on input attributes. The output classifications for an ML model trained for evidence classifications 136 may be classifications of the categories and/or codes for uncategorized evidence. The output classifications for an ML model trained for evidence recommendations 138 may be classifications of the categories and/or types of evidence recommended for a dispute or dispute type (e.g., based on dispute details or attributes) from win probabilities of evidence classifications 136. Evidence classifications 136 may include evidence previously classified into categories, such as by a data scientist, administrator, machine or other entity that classifies the evidence to categories or otherwise annotates the training data for evidence classification. Evidence classifications 136 may also include the uncategorized data now classified into categories by one or more of ML models 134. Thereafter, evidence classifications 136 may be used to further train one or more of ML models 134, which the provides evidence recommendations 138 based on received disputes, ¶44; At step 406, a first ML model for evidence classification is trained using the extracted text data. The first ML model may be trained using attributes from the evidence classification training data, including those attributes identified from the extracted text data (e.g., words, phrases, and the like that may be converted to vectors for vector comparisons and the like when training layers of a ML model). Further, the input attributes may include the evidence categories corresponding to the extracted text data. For example, the evidence in the training data may be classified as categories or types corresponding to codes or identifiers such that the extracted text data may then be correlated to these categories. This data therefore allows training of the first ML model to make classifications of text from other evidence into categories from the training data, ¶72).
As per claim 6, Ramanjani and Ross disclose as shown above with respect to claim 4. Ross further teaches generating, by applying at least one weighted matrix and at least one activation function at the input layer, the plurality of hidden layers, and the output layer, the at least one machine learning score from the input vector (In some examples, prediction models may cluster points of data in functional groups such as “random forests.” Random Forests may comprise combinations of decision tree predictors (input layers, hidden layers, ¶41-¶43; Thus, the training data may be used as input/output data sets that allow for ML models 134 to make classifications based on input attributes. The output classifications for an ML model trained for evidence classifications 136 may be classifications of the categories and/or codes for uncategorized evidence. The output classifications for an ML model trained for evidence recommendations 138 may be classifications of the categories and/or types of evidence recommended for a dispute or dispute type (e.g., based on dispute details or attributes) from win probabilities of evidence classifications 136. Evidence classifications 136 may include evidence previously classified into categories, such as by a data scientist, administrator, machine or other entity that classifies the evidence to categories or otherwise annotates the training data for evidence classification. Evidence classifications 136 may also include the uncategorized data now classified into categories by one or more of ML models 134. Thereafter, evidence classifications 136 may be used to further train one or more of ML models 134, which the provides evidence recommendations 138 based on received disputes, ¶44; At step 406, a first ML model for evidence classification is trained using the extracted text data. The first ML model may be trained using attributes from the evidence classification training data, including those attributes identified from the extracted text data (e.g., words, phrases, and the like that may be converted to vectors for vector comparisons and the like when training layers of a ML model). Further, the input attributes may include the evidence categories corresponding to the extracted text data. For example, the evidence in the training data may be classified as categories or types corresponding to codes or identifiers such that the extracted text data may then be correlated to these categories. This data therefore allows training of the first ML model to make classifications of text from other evidence into categories from the training data, ¶72).
As per claim 9, Ramanjani and Ross disclose as shown above with respect to claim 1. Ramanjani further discloses determining a predefined time interval for submitting the contestation document to contest the chargeback request; and based on the determining, activating the machine learning framework to generate the at least one machine learning score on demand or at predefined time intervals (In one embodiment, recommendation module 307 may generate a recommended action in response to the disputed transaction, e.g., chargeback based on the data received from machine learning module 207 and prediction module 209. In another embodiment, recommendation module 307 may collaborate with user interface module 211 to generate a representment notification to inform the users of the decision to dispute the chargeback or accept financial liability. In one example embodiment, recommendation module 307 may determine that merchants have a higher probability of winning chargeback representments if the dispute is filed within a pre-defined time period, e.g., 14 days, based on historical data. Recommendation module 307 may recommend a merchant to dispute the chargeback within the pre-defined time period, Ramanjani ¶63).
As per claim 16, Ramanjani and Ross disclose as shown above with respect to claim 10. Ramanjani further discloses wherein the data associated with the transaction in the contestation document includes transaction data and service provider data (FIG. 7 is a schematic diagram illustrating a user interacting with an electronic device comprising a user interface that represents a presentation of prediction, probability, and recommendation on chargeback representment, according to one example embodiment. In one embodiment, user interface 701 includes transaction 703, prediction 705, win probability 707, loss probability 709, and recommendation 711 for a disputed transaction. In one embodiment, transaction 703 includes transaction information, e.g., transaction number, data of transaction, transaction value, etc. The user can access additional information by hovering the cursor over transaction 703 or by clicking transaction 703. In one embodiment, prediction 705 represents the prediction results based on the calculation by prediction module 209. In one embodiment, win probability 707 and loss probability 709 are based on the calculation by machine learning module 207. For example, win probability 707 indicates the probability a merchant will win a chargeback representment and loss probability 709 indicates the probability a merchant will lose a chargeback representment. In one embodiment, recommendation 711 advises a user to either accept financial liability or proceed with chargeback representment. Such advice is based on the prediction data and the probability data. In one example embodiment, transaction that satisfies the minimum threshold value and the minimum threshold probability requirement are recommended for chargeback representment, Ramanjani Fig. 7 and ¶91).
As per claim 18, Ramanjani and Ross disclose as shown above with respect to claim 17. Ramanjani further discloses determining that a time to respond to the chargeback request is less than a predefined time; and activating the machine learning model to generate the machine learning score when the time to respond to the chargeback request is less than the predefined time (In one embodiment, recommendation module 307 may generate a recommended action in response to the disputed transaction, e.g., chargeback based on the data received from machine learning module 207 and prediction module 209. In another embodiment, recommendation module 307 may collaborate with user interface module 211 to generate a representment notification to inform the users of the decision to dispute the chargeback or accept financial liability. In one example embodiment, recommendation module 307 may determine that merchants have a higher probability of winning chargeback representments if the dispute is filed within a pre-defined time period, e.g., 14 days, based on historical data. Recommendation module 307 may recommend a merchant to dispute the chargeback within the pre-defined time period, Ramanjani ¶63).
As per claim 19, Ramanjani and Ross disclose as shown above with respect to claim 17. Ramanjani further discloses training the machine learning model using historical data associated with historical transactions, wherein historical transactions include disputed transactions and not disputed transactions and is collected over a predefined time period (training the module during the training phase, with historical information associated with user, Ramanjani ¶51).
As per claim 21, Ramanjani and Ross disclose as shown above with respect to claim 1. Ramanjani further discloses wherein the historical data comprises device data from which one or more transactions have occurred over the predefined time period (In one embodiment, probability module 305 may implement machine-learned rules to calculate the likelihood a user may win a chargeback representment. The calculation may include determining a win probability based, at least in part, on an analysis of data of the actual transaction and historical big data, e.g., data of past transactions. In one embodiment, historical data may include customer-specific information indicative of how often a specific customer successfully or unsuccessfully disputed online transactions. In another embodiment, historical data may include merchant-specific information indicative of how often a specific merchant successfully disputed particular types of chargebacks. In a further embodiment, historical data may include global historical win rates (or percentages) indicative of how often, based on the particular chargeback reason code, a group of merchants, e.g., a global population of merchants, successfully disputed particular types of chargebacks, Ramanjani ¶60).
As per claim 22, Ramanjani and Ross disclose as shown above with respect to claim 21. Ramanjani further discloses wherein the device data comprises one or more of an Internet Protocol address, a device identifier, or a biometric identifier (device information, MEID, Ramanjani ¶49; IP based network protocols, ¶111).
Claim(s) 8, 13-15, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ramanjani et al. (US PG Pub. 2023/0281635) and Ross et al. (US PG Pub. 2022/0309507) further in view of Fenichel et al. (US PG Pub. 2024/0070671).
As per claim 8, Ramanjani and Ross disclose as shown above with respect to claim 1. The combination of Ramanjani and Ross do not expressly disclose wherein the at least one machine learning model includes an ensemble of trees, and further comprising: propagating the input vector through each tree in the ensemble of trees by comparing variables in the input vector to variables at nodes in the each tree until a leaf node of the each tree is reached; determining a tree score for the each tree based on the leaf node of the each tree; and combining tree scores of the ensemble of trees into the at least one machine learning score
Howver, Fenichel teaches wherein the at least one machine learning model includes an ensemble of trees, and further comprising: propagating the input vector through each tree in the ensemble of trees by comparing variables in the input vector to variables at nodes in the each tree until a leaf node of the each tree is reached; determining a tree score for the each tree based on the leaf node of the each tree; and combining tree scores of the ensemble of trees into the at least one machine learning score (In some examples, prediction models may cluster points of data in functional groups such as “random forests.” Random Forests may comprise combinations of decision tree predictors. (Decision trees may comprise a data structure mapping observations about something, in the “branch” of the tree, to conclusions about that thing's target value, in the “leaves” of the tree.) Each tree may depend on the values of a random vector sampled independently and with the same distribution for all trees in the forest. Prediction models may also include artificial neural networks. Artificial neural networks may model input/output relationships of variables and parameters by generating a number of interconnected nodes which contain an activation function. The activation function of a node may define a resulting output of that node given an argument or a set of arguments. Artificial neural networks may generate patterns to the network via an ‘input layer’, which communicates to one or more “hidden layers” where the system determines regressions via a weighted connections. Prediction models may additionally or alternatively include classification and regression trees, or other types of models known to those skilled in the art. To generate prediction models, the fraud detection system may analyze information applying machine-learning methods, Fenichel ¶56).
The Fenichel, Ross, and Ramanjani references are analogous in that both are directed towards/concerned with dispute resolution. Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to use Fenichel’s ability to utilize tree structures in Ross and Ramanjani’s system to improve the system and method with reasonable expectation that this would result in a dispute resolution management system that is able to process more evidence for merchants.
The motivation being that there is a need for improved systems and methods for detecting fraudulent activity (Fenichel ¶3).
As per claim 13, Ramanjani and Ross disclose as shown above with respect to claim 10. While Ramanjani further discloses the use of convolutional neural networks (Ramanjani ¶94) and layers thereof (Ramanjani ¶95), the combination of Ramanjani and Ross do not expressly disclose wherein the at least one machine learning model is a convolutional neural network model with a weighted matrix and an activation function at each neuron in the plurality of hidden layers.
However, Fenichel teaches wherein the at least one machine learning model is a convolutional neural network model including a plurality of hidden layers with a weighted matrix and an activation function at each neuron in the plurality of hidden layers (In some examples, prediction models may cluster points of data in functional groups such as “random forests.” Random Forests may comprise combinations of decision tree predictors. (Decision trees may comprise a data structure mapping observations about something, in the “branch” of the tree, to conclusions about that thing's target value, in the “leaves” of the tree.) Each tree may depend on the values of a random vector sampled independently and with the same distribution for all trees in the forest. Prediction models may also include artificial neural networks. Artificial neural networks may model input/output relationships of variables and parameters by generating a number of interconnected nodes which contain an activation function. The activation function of a node may define a resulting output of that node given an argument or a set of arguments. Artificial neural networks may generate patterns to the network via an ‘input layer’, which communicates to one or more “hidden layers” where the system determines regressions via a weighted connections. Prediction models may additionally or alternatively include classification and regression trees, or other types of models known to those skilled in the art. To generate prediction models, the fraud detection system may analyze information applying machine-learning methods, Fenichel ¶56).
The Fenichel, Ross, and Ramanjani references are analogous in that both are directed towards/concerned with dispute resolution. Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to use Fenichel’s ability to have layers in machine learning algorithms in Ramanjani’s system to improve the system and method with reasonable expectation that this would result in a dispute resolution management system that is able to process more evidence for merchants.
The motivation being that there is a need for improved systems and methods for detecting fraudulent activity (Fenichel ¶3).
As per claim 14, Ramanjani, Ross, and Fenichel disclose as shown above with respect to claim 13. Fenichel further teaches applying, at each neuron in a hidden layer of the plurality of hidden layers, the weighted matrix and the activation function to each input layer output, to generate a hidden layer output; and generating, at the output layer of the machine learning model, the machine learning score from the hidden layer output (Decision trees may comprise a data structure mapping observations about something, in the “branch” of the tree, to conclusions about that thing's target value, in the “leaves” of the tree.) Each tree may depend on the values of a random vector sampled independently and with the same distribution for all trees in the forest. Prediction models may also include artificial neural networks. Artificial neural networks may model input/output relationships of variables and parameters by generating a number of interconnected nodes which contain an activation function. The activation function of a node may define a resulting output of that node given an argument or a set of arguments. Artificial neural networks may generate patterns to the network via an ‘input layer’, which communicates to one or more “hidden layers” where the system determines regressions via a weighted connections. Prediction models may additionally or alternatively include classification and regression trees, or other types of models known to those skilled in the art. To generate prediction models, the fraud detection system may analyze information applying machine-learning methods, Fenichel ¶56).
As per claims 15 and 20, Ramanjani and Ross discloses as shown above with respect to claims 10 and 17. While Ramanjani further discloses the use of decision trees (Ramanjani ¶52), the combination of Ramanjani and Ross do not expressly disclose wherein the at least one machine learning model includes an ensemble of trees, and further comprising: propagating the input vector through each tree in the ensemble of trees by comparing variables in the input vector to variables at nodes in each tree until a leaf node of each tree is reached; determining a tree score for each tree based on the leaf node of each tree; applying a weight to the tree score for each tree; and combining weighted tree scores of the ensemble of trees into the machine learning score; training the machine learning model to generate an ensemble of trees using historical data associated with the transactions wherein nodes of each tree in the ensemble of trees includes at least one variable and a value associated with the at least one variable that determine a path through the each tree, wherein the path determines the machine learning score.
However, Fenichel teaches wherein the at least one machine learning model includes an ensemble of trees, and further comprising: propagating the input vector through each tree in the ensemble of trees by comparing variables in the input vector to variables at nodes in each tree until a leaf node of each tree is reached; determining a tree score for each tree based on the leaf node of each tree; applying a weight to the tree score for each tree; and combining weighted tree scores of the ensemble of trees into the machine learning score; training the machine learning model to generate an ensemble of trees using historical data associated with the transactions wherein nodes of each tree in the ensemble of trees includes at least one variable and a value associated with the at least one variable that determine a path through the each tree, wherein the path determines the machine learning score (Decision trees may comprise a data structure mapping observations about something, in the “branch” of the tree, to conclusions about that thing's target value, in the “leaves” of the tree.) Each tree may depend on the values of a random vector sampled independently and with the same distribution for all trees in the forest. Prediction models may also include artificial neural networks. Artificial neural networks may model input/output relationships of variables and parameters by generating a number of interconnected nodes which contain an activation function. The activation function of a node may define a resulting output of that node given an argument or a set of arguments. Artificial neural networks may generate patterns to the network via an ‘input layer’, which communicates to one or more “hidden layers” where the system determines regressions via a weighted connections. Prediction models may additionally or alternatively include classification and regression trees, or other types of models known to those skilled in the art. To generate prediction models, the fraud detection system may analyze information applying machine-learning methods, Fenichel ¶56).
The Fenichel, Ross, and Ramanjani references are analogous in that both are directed towards/concerned with dispute resolution. Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to use Fenichel’s ability to have layers in machine learning algorithms in Ramanjani’s system to improve the system and method with reasonable expectation that this would result in a dispute resolution management system that is able to process more evidence for merchants.
The motivation being that there is a need for improved systems and methods for detecting fraudulent activity (Fenichel ¶3).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ANDREW B WHITAKER whose telephone number is (571)270-7563. The examiner can normally be reached on M-F, 8am-5pm, EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Lynda Jasmin can be reached on (571) 272-6782. 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 the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ANDREW B WHITAKER/Primary Examiner, Art Unit 3629
