DETAILED ACTION
Status of Claims
The present application is being examined under the AIA first to file provisions.
This action is in reply to the amendment filed on 06/16/2026.
Claims 1-3, 15, 29, and 33 have been amended and are hereby entered.
Claims 1-36 are currently pending and have been examined.
Response to Arguments
Applicant's arguments filed 06/16/2026 with respect to the 101 rejection for the claims being directed towards an abstract idea have been fully considered but they are not persuasive.
Applicant argues #1:
A. The claims do not recite certain methods of organizing human activity
Viewed as a whole, the claims are directed to the operation of a computer system-capturing transaction data, converting document images into structured and verified machine-readable records, training and applying a neural network, and rendering and updating a graphical user interface-rather than to interactions between people. The claims do not set forth or describe a fundamental economic practice, an advertising, marketing, or sales activity, or the management of relationships, agreements, or legal obligations between parties, and no claimed step is performed by or between parties to a commercial transaction. Although the claimed system operates on financial data and its output relates to credit, the financial subject matter is the data environment in which the recited machine operations execute; a claim does not recite commercial or legal interactions merely because the data it processes is financial in character. See MPEP § 2106.04(a)(2)(II).
Examiners response:
The Examiner respectfully disagrees, while the claims do recite additional elements of the computer system and neural network, the focus of the claims are directed towards determining creditworthiness for users in emerging markets who lack formal credit histories and analyzing transactions to do so, which under broadest reasonable interpretation describes commercial and legal interactions for an abstract way to analyze information for determining a credit score.
For the reasons above, applicant’s arguments are not persuasive.
Applicant argues #2:
B. The claims do not recite mathematical concepts
The claims do not set forth or describe a mathematical relationship, mathematical formula or equation, or mathematical calculation. No claim limitation recites a formula or requires that a computation be expressed in mathematical terms; the claims instead recite operations of a computer system that, like virtually all computer-implemented processes, are ultimately based on mathematics. A claim that merely involves or is based on a mathematical concept does not recite one. MPEP § 2106.04(a)(2)(I). The USPTO's August 4, 2025 Memorandum on evaluating subject matter eligibility confirms, for example, that the training of a neural network does not per se recite a judicial exception merely because the training involves mathematical operations.
Examiners response:
The Examiner respectfully disagrees, with regards to the mathematical concepts, as an initial matter, the Examiner did not rely on grouping the claims into the mathematical grouping of abstract ideas for the analysis. Furthermore Examiners are directed to continue to use the Mayo Alice framework (as laid out in MPEP 2106 which incorporates Steps 2A and Step 2B of the 2019 PEG) as guidance in evaluating subject matter eligibility, which the Examiner has properly applied. Therefor this argument is not persuasive.
Applicant argues #3:
C. The claims do not recite mental processes
Nor do the claims recite concepts that can practically be performed in the human mind. The claims as a whole require operations that can only be performed by a computer, including parsing electronic images with optical character recognition and named-entity-recognition components, propagating observations through a multi-layer neural network, updating network weights across batches of observations and over a plurality of epochs, and automatically retraining and rescoring upon each newly captured transaction. Under the USPTO's 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence (89 FR 58128 (July 17, 2024)), and the August 4, 2025 Memorandum, the mental process grouping "is not without limits" and does not encompass claim limitations that cannot practically be performed in the human mind. See MPEP § 2106.04(a)(2)(III).
Because the claims, read as a whole, do not set forth or describe an abstract idea within any of the enumerated groupings, the claims do not recite a judicial exception and are patent eligible at Step 2A, Prong One. In the alternative, the claims integrate any alleged exception into a practical application, as set forth below.
Examiners response:
The Examiner respectfully disagrees, with regards to the mental processes, as an initial matter, the Examiner did not rely on grouping the claims into the mental processes grouping of abstract ideas for the analysis and just because an idea cannot be performed with a pen and paper or in the human mind does not mean it' s not directed towards an abstract idea. Furthermore Examiners are directed to continue to use the Mayo Alice framework (as laid out in MPEP 2106 which incorporates Steps 2A and Step 2B of the 2019 PEG) as guidance in evaluating subject matter eligibility, which the Examiner has properly applied. Therefor this argument is not persuasive.
Therefor this argument is not persuasive.
Applicant argues #4:
IV. Step 2A, Prong Two -- The Claims Integrate Any Alleged Abstract Idea Into a
Practical Application
A. The claims recite a specific technical solution to a technical problem of data capture
The problem addressed by the amended claims is a technical problem of data capture. The specification explains that when a user pays a bill outside the system, transaction data can be ingested automatically only if "an external transaction account may be linked to the service for the purpose of ingesting transaction data," in which case "data is fetched from the payment provider and is parsed" (176). When no transaction account is linked, however, no machine-readable record of the transaction is available to the system, and the only record of the transaction may be a document (see 177). An uploaded proof of payment is unstructured image data that a credit-scoring computation cannot consume directly, and the specification further explains that such a transaction must be screened before it can be trusted: "certain transactions may be verified as legitimate prior to being used for adjusting a person's credit score" ( 76).
The present application solves this problem with a particular parsing-and-verification pipeline: "the system may receive a proof of payment which is parsed by an Optical Character Recognition/Intelligent Character Recognition module and Named Entity Recognition service. Such information may be used to verify the transactions and prevent fraud" (Q 77). The amended claims recite this solution:
"receiving, at the transactions processing service from a user device, an electronic image of a proof of payment...; parsing the electronic image using an optical character recognition or intelligent character recognition module and a named entity recognition service to extract a structured transaction record comprising the identification of the purchased good or purchased service and the price for the purchased good or purchased service; and executing a transaction verification and fraud protection process on the structured transaction record to produce verified transaction information."
Through these limitations, the claimed system goes beyond receiving or transmitting data over a network: it computationally generates structured, verified, machine-readable transaction data from an unstructured document image, where no such data previously existed, and gates that data through a fraud-protection process before it is stored or used to train the recited neural network. That is a specific means or method that solves a problem in an existing technological process, see Koninklijke KPN NV v. Gemalto M2M GmbH, 942 F.3d 1143, 1150 (Fed. Cir. 2019), and a solution "necessarily rooted in computer technology," DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1257 (Fed. Cir. 2014). The problem the claims address, and the manner in which they solve it, are accordingly technical in nature.
Examiners response:
The Examiner respectfully disagrees, uploading/OCR’ing proof of payments for verifying transactions is not indicative of an improvement in technology, rather this is merely using the additional as tools for performing record keeping task akin to Electronically scanning or extracting data from a physical document, Content Extraction and Transmission, LLC v. Wells Fargo Bank, 776 F.3d 1343, 1348, 113 USPQ2d 1354, 1358 (Fed. Cir. 2014) (optical character recognition); and Electronic recordkeeping, Alice Corp. Pty. Ltd. v. CLS Bank Int'l, 573 U.S. 208, 225, 110 USPQ2d 1984 (2014) (creating and maintaining "shadow accounts"); Ultramercial, 772 F.3d at 716, 112 USPQ2d at 1755 (updating an activity log); in which the courts found these functions to be WURC. Further, the claims here are not like those the Court found patent eligible in DDR, in which the inventive concept was in the modification of conventional mechanics behind website display to produce a dual-source integrated hybrid display because applicant’s claims here do not address problems unique to the Internet or require an arguably inventive device or technique for displaying information and further not akin to Koninklijke where the Courts found the claims to be patent-eligible because they were directed to a non-abstract improvement in an existing technological process (i.e., error checking in data transmissions by varying the way check data is generated to achieve this increased detection capability. )
Applicant argues #5:
B. The claims delineate how the machine learning technology achieves the result
The amended claims also recite the specific manner in which the machine learning model is constructed and trained, by reciting a particular network topology:
"the neural network having an input layer comprising a number of input nodes equal to a number of the plurality of independent variables, at least one hidden layer applying a rectifier activation function, and an output neuron applying a sigmoid activation function to produce a continuous-value output representing the individually-determined credit score"
and a particular training procedure:
"standardiz[ing] [each independent variable] to a mean of zero and a variance of one," "forward-propagat[ing] an observation of a training data set through the neural network to produce a predicted result," "evaluat[ing] a gradient descent cost function ... by comparing the predicted result to an actual result,"
and "back-propagating the error via stochastic gradient descent, the one or more weights being updated only after a batch of observations of the training data set has been forward-propagated, over a plurality of epochs as additional verified transaction information is acquired."
These limitations place the amended claims within the space that Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025), expressly preserves for machine learning inventions. The Federal Circuit held "only that patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101," id. at 1214, and confirmed that "machine learning is a burgeoning and increasingly important field and may lead to patent-eligible improvements in technology," id. at 1213. The Recentive claims were found ineligible because they described what the model produced (optimized schedules; network maps) without delineating "the steps through which the machine learning technology achieves an improvement." Id. The amended claims here recite those very steps-the specific architecture and the specific training methodology through which the result is achieved-in combination with the parsing-and-verification pipeline described above that supplies the model's inputs. The claims thus do considerably more than apply generic machine learning to a new data environment.
Post-Recentive decisions confirm this distinction. In Aon Re, Inc. v. Zesty.AI, Inc., No. CV 25- 201, 2025 WL 1938214 (D. Del. July 15, 2025), the court sustained eligibility where the claim "implements machine-learning technology in a specific way to address a practical technical problem." In Nielsen Co. (US), LLC v. Hyphametrics, Inc., No. CV 23-136-GBW, 2025 WL 1672002 (D. Del. June 13, 2025), the court sustained eligibility where the claims provided "an improvement to the technical field of image processing" and "recite[d] the specific [neural] networks used to perform the detection." The amended claims likewise recite the specific network and procedure used-and, like Nielsen, include a specific image-processing component (the OCR/ICR and named-entity-recognition parsing limitation).
The same considerations distinguish the amended claims from claim 2 of Example 47 of the July 2024 Subject Matter Eligibility Examples. Whereas that claim recited training a neural network without any particular architecture or methodology, amended claim 1 specifies the topology of the network, the activation functions applied at each layer, the form of the output, the standardization of the inputs, and the stochastic-gradient-descent regimen by which the weights are updated. These recited specifics provide the very detail whose absence rendered the example claim ineligible.
Examiners response:
The Examiner respectfully disagrees, with respect to the neural network and standardizing the data and back-propagating the error via stochastic gradient descent, these limitations fail to render the claims eligible as they are further limiting the abstract idea, and akin to Recentive Analytics, Inc. v. Fox Corp., Case No. 2023-2437 (Fed. Cir. Apr. 18, 2025), where the Courts found that instead of disclosing “a specific implementation of a solution to a problem in the software arts,” Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1339 (Fed. Cir. 2016), or “a specific means or method that solves a problem in an existing technological process,” Koninklijke, 942 F.3d at 1150, the only thing the claims disclose about the use of machine learning is that machine learning is used in a new environment and that the requirements that the machine learning model be “iteratively trained” or dynamically adjusted in the Machine Learning Training patents was not a technological improvement in that iterative training using selected training material and dynamic adjustments based on real-time changes are incident to the very nature of machine learning.
In response the argument regarding the decisions of Aon Re, Inc. v. Zesty.AI, Inc., No. CV 25- 201, 2025 WL 1938214 (D. Del. July 15, 2025), and Nielsen Co. (US), LLC v. Hyphametrics, Inc., No. CV 23-136-GBW, 2025 WL 1672002 (D. Del. June 13, 2025) and that that the court sustained eligibility under 35 USC 101, these Court decisions are not precedential and do not represent office policy and moreover, those decisions were fact specific to the cases being decided and as discussed above, OCR’ing a document is WURC. As such, the claims have been fully analyzed under the 101 guidance in MPEP 2106 and in accordance with Office policies and the Examiner maintains the 101 rejection.
Applicant argues #6:
C. The claims recite a specific, non-generic distributed architecture
The Office Action (pp. 14-16) characterized additional elements ("a memory unit, a computer-readable storage media, one or more processers [sic], processing module, an API gateway, display for a graphical user interface, and machine learning model/unit") as "recited at a high level of generality." The amended claims now recite a particular distributed arrangement of these elements:
"a plurality of containerized microservices comprising at least a transactions processing service, a credit scoring service, and a pricing service, each exposing an internal application programming interface (API), the plurality of containerized microservices being accessed through an API gateway comprising a load balancer and a service register that routes requests to respective ones of the plurality of containerized microservices."
In Amdocs (Israel) Ltd. v. Openet Telecom, Inc., 841 F.3d 1288, 1300-01 (Fed. Cir. 2016), the Federal Circuit sustained eligibility for claims requiring a distributed architecture of components working in conjunction, explaining that such claims are "unlike the claims in Electric Power Group, which failed to pass muster because they simply gathered and analyzed information using generic computer components in a conventional way." The amended claims align with the Amdocs side of that line: they recite discrete, enumerated services with defined interfaces, arranged behind a structurally defined gateway, performing a defined sequence of operations.
Examiners response:
The Examiner respectfully disagrees, further including the load balancer and API gateway does not render the claims eligible as this is generally linking the idea to the computer environment. And as per MPEP 2106.05(f) Use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application or provide significantly more. See Affinity Labs v. DirecTV, 838 F.3d 1253, 1262, 120 USPQ2d 1201, 1207 (Fed. Cir. 2016) (cellular telephone);. In Amdocs, the claims are directed towards the network processing data in way that was unconventional resulting to an improvement to the technical area for processing data by reducing the flow of data records and reducing the amount of data stored in the central database, thus eliminating capacity bottlenecks, and improving the scalability and efficiency of the system, resulting in an improvement to the additional elements themselves, unlike Amdocs there is no improvement the additional elements themselves, the load balancer and API gateway are generically claimed such that the claims are further describing the technical environment in which the idea is being limited to.
Applicant argues #7:
D. Considered as a whole, the claims recite a particular solution, not the idea of a
solution
The USPTO's August 4, 2025 Memorandum instructs examiners to evaluate "all the claim limitations and how these limitations interact and impact each other," and to consider whether the claim covers "a particular solution to a problem or a particular way to achieve a desired outcome." The amended claims recite an integrated, end-to-end pipeline: an electronic image of a proof of payment is received at a recited microservice; the image is parsed by OCR/ICR and named-entity- recognition components into a structured transaction record; the record passes a fraud-protection verification before storage; the verified data is standardized and used to train a neural network of recited topology by a recited stochastic-gradient-descent procedure; the trained model determines the credit score; the score drives the determination of "one or more first terms" and "one or more second terms...including price and credit terms"; the terms are rendered in "a display for a graphical user interface"; and, "responsive to capturing and categorizing a new financial transaction made by the user," the system "automatically retrain[s] the machine learning model, recalculate[s] the individually-determined credit score, re-determine[s] the one or more first terms and the one or more second terms, and update[s] the display of the graphical user interface." Each limitation supplies the operand for the next. That is a particular way of achieving the outcome-not a claim to the outcome itself-and it integrates any alleged abstract idea into a practical application. The rejection should be withdrawn at Step 2A, Prong Two, without reaching Step 2B.
V. Step 2B - In the Alternative, the Claims Recite Significantly More Than Any Alleged
Abstract Idea
A. The ordered combination is non-conventional
Even if the analysis reached Step 2B (which Applicant respectfully submits it should not), "an inventive concept can be found in the non-conventional and non-generic arrangement of known, conventional pieces." BASCOM Global Internet Svcs., Inc. v. AT&T Mobility LLC, 827 F.3d 1341, 1350 (Fed. Cir. 2016). Here, the claimed arrangement-an OCR/ICR and named-entity-recognition extraction stage that creates machine-readable records for transactions "having no machine-readable record accessible to the system," a "transaction verification and fraud protection process" that gates which records may be stored and used for training, a neural network of recited topology trained by batch stochastic gradient descent on the verified records, and automatic retraining and re- determination of terms "responsive to capturing and categorizing a new financial transaction"-is precisely such an ordered combination. This particular combination of elements, considered together, supplies an inventive concept independent of the eligibility of any element viewed in isolation.
B. The Office Action provides no evidence that the claimed combination is well- understood, routine, and conventional
"Whether something is well-understood, routine, and conventional to a skilled artisan at the time of the patent is a factual determination" that must be supported by evidence. Berkheimer v. HP Inc., 881 F.3d 1360, 1369 (Fed. Cir. 2018); MPEP § 2106.05(d)(I) (requiring a citation to an express specification admission, a court decision noted in MPEP § 2106.05(d)(II), a publication, or official notice). The authorities cited (pp. 16-17) in the Office Action address receiving and transmitting data over a network, performing repetitive calculations, and updating an activity log. The amended claims recite operations of a different character-parsing a document image with OCR/ICR and named- entity-recognition components to generate verified structured transaction records, a neural network of recited topology trained by batch stochastic gradient descent, and the combination of those elements within the recited microservice architecture. Applicant respectfully submits that, on the present record, these elements provide significantly more than any alleged abstract idea.
Examiners response:
The Examiner respectfully disagrees, for the same reasons as discussed above with respect to the training/retraining the neural network and parsing the images to verify the transactions, the claims do not recite significantly more than the abstract idea. Further, even though the claims are particular in that they overcome the art of record, just because claims may be novel under § 103 over a number of prior art rejections does not mean they are not directed to an abstract idea. Cf. Intellectual Ventures ILLCv. Symantec Corp., 838 F.3d 1307, 1315 (Fed. Cir. 2016).
Indeed, “[t]he ‘novelty’ of any element or steps in a process, or even of the process itself, is of no relevance in determining whether the subject matter of a claim falls within the § 101 categories of possibly patentable subject matter.” Diamond v. Diehr, 450 U.S. 175, 188—89 (1981) (emphasis added); see also Mayo, 132 S. Ct. at 1303—04 (rejecting “the Government’s invitation to substitute §§ 102, 103, and 112 inquiries for the better established inquiry under § 101”). Here, the jury’s general finding that Symantec did not prove by clear and convincing evidence that three particular prior art references do not disclose all the limitations of or render obvious the asserted claims does not resolve the question of whether the claims embody an inventive concept at the second step of Mayo/Alice.
Additionally, the claims here are not like those the Court found patent eligible in Bascom, in which the inventive concept was the unconventional arrangement of the installation of a filtering tool at a specific location, remote from the end-users, with customizable filtering features specific to each end user, this design permitted the filtering tool to have both the benefits of a filter on a local computer and the benefits of a filter on the [Internet Service Provider] server and was not conventional or generic, instead, the patent claimed and explained how a particular arrangement of elements was “a technical improvement over prior art ways of filtering such content.” (BASCOM, 827 F.3d at 1345.). In the instant application the claims do not have an inventive concept found in the non-conventional and non-generic arrangement of the additional elements and for the reasons discussed above fail to amount to a technical improvement or significantly more, even when considered in combination and as a whole.
For the reasons above, the 101 rejection of claims 1-36 is hereby maintained.
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.
Subject Matter Eligibility Test under 101
Claims 1-36 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more, and fails step 2 of the analysis because the focus of the claims is not on the devices themselves or a practical application but rather directed towards an abstract idea, the analysis is provided below.
Step 1 (Statutory Categories) – The claims pass step 1 of the subject matter eligibility test (see MPEP 2106(III)) as the claims are directed towards a system and methods.
Step 2A – Prong One (Do the claims recite an abstract idea?) -
Claims 1 recites an idea, in part, by:
create a digital profile for a user, the digital profile being used by the user in making one or more financial transactions;
track transaction information of the user based on the digital profile, the transaction information relating to financial transactions made by the user comprising informal economic activities including at least one of rent payments, utility payments, and bill payments that are not formally recorded in traditional credit bureau systems, at least one of the financial transactions comprising an identification of a purchased good or purchased service, and a price for the purchased good or purchased service, wherein tracking the transaction information comprises:
receiving, at the transactions processing service from a user, an electronic image of a proof of payment for a first one of the financial transactions, the first one of the financial transactions having no machine-readable record accessible to the system;
parsing the electronic image using an optical character recognition or intelligent character recognition module and a named entity recognition service to extract a structured transaction record comprising the identification of the purchased good or purchased service and the price for the purchased good or purchased service; and
executing a transaction verification and fraud protection process on the structured transaction record to produce verified transaction information; 5
store the verified transaction information relating to the financial transactions made by the user with reference to the digital profile;
train a machine learning model configured to determine, for the user, an individually-determined credit score for the user based on, at least in part, the verified transaction information relating to the financial transactions made by the user, wherein training the machine learning model comprises:
initializing a neural network comprising at least one differential equation and a plurality of neurons having a plurality of independent variables derived from the verified transaction information, each standardized to a mean of zero and a variance of one, as one or more inputs, and one or more weights;
forward-propagating an observation of a training data set through the neural network to produce a predicted result, wherein a value computed by a first one of the plurality of neurons is input to a second one of the plurality of neurons;
evaluating a gradient descent cost function to determine a value for an error of the neural network by comparing the predicted result to an actual result; and
adjusting values of the one or more weights to reduce the error of the neural network by back-propagating the error via stochastic gradient descent, the one or more weights being updated only after a batch of observations of the training data set has been forward-propagated, over a plurality of epochs as additional verified transaction information is acquired;
apply the machine learning model to the verified transaction information relating to the financial transactions made by the user to determine the individually-determined credit score for the user;
determine interests for the user in a bundle of a plurality of goods or services based on the verified transaction information; and
determine one or more first terms for at least a first one of the bundle of the plurality of the goods or services based on the individually-determined credit score and one or more second terms for at least a second one of the bundle of the plurality of the goods or services, the one or more first terms and the one or more second terms including price and credit terms; and
generate for a user the one or more first terms for the at least first one of the bundle of the plurality of goods or services and the one or more second terms for the at least second one of the bundle of the plurality of goods or services; and
responsive to capturing and categorizing a new financial transaction made by the user, automatically retrain the machine learning model, recalculate the individually-determined credit score, re-determine the one or more first terms and the one or more second terms.
Claims 15 recites an idea, in part, by:
create a digital profile for a user, the digital profile being used by the user in making one or more financial transactions;
track transaction information of the user based on the digital profile, the transaction information relating to financial transactions made by the user comprising informal economic activities including at least one of rent payments, utility payments, and bill payments that are not formally recorded in traditional credit bureau systems, at least one of the financial transactions comprising an identification of a purchased good or purchased service, and a price for the purchased good or purchased service, wherein tracking the transaction information comprises:
receiving, at the transactions processing service from a user, an electronic image of a proof of payment for a first one of the financial transactions, the first one of the financial transactions having no machine-readable record accessible to the system;
parsing the electronic image using an optical character recognition or intelligent character recognition module and a named entity recognition service to extract a structured transaction record comprising the identification of the purchased good or purchased service and the price for the purchased good or purchased service; and
executing a transaction verification and fraud protection process on the structured transaction record to produce verified transaction information; 5
storing the verified transaction information relating to the financial transactions made by the user with reference to the digital profile;
training a machine learning model configured to determine, for the user, an individually-determined credit score for the user based on, at least in part, the verified transaction information relating to the financial transactions made by the user, wherein training the machine learning model comprises:
initializing a neural network comprising at least one differential equation and a plurality of neurons having a plurality of independent variables derived from the verified transaction information, each standardized to a mean of zero and a variance of one, as one or more inputs and one or more weights, the neural network having an input layer comprising a number of input nodes equal to a number of the plurality of independent variables, at least one hidden layer applying a rectifier activation function, and an output neuron applying a sigmoid activation function to produce a continuous-value output representing the individually-determined credit score;
forward-propagating an observation of a training data set through the neural network to produce a predicted result, wherein a value computed by a first one of the plurality of neurons is input to a second one of the plurality of neurons;
evaluating a gradient descent cost function to determine a value for an error of the neural network by comparing the predicted result to an actual result; and
adjusting values of the one or more weights to reduce the error of the neural network by back-propagating the error via stochastic gradient descent, the one or more weights being updated only after a batch of observations of the training data set has been forward-propagated, over a plurality of epochs as additional verified transaction information is acquired;
apply the machine learning model to the verified transaction information relating to the financial transactions made by the user to determine the individually-determined credit score for the user;
determine interests for the user in a bundle of a plurality of goods or services based on the verified transaction information; and
determine one or more first terms for at least a first one of the bundle of the plurality of the goods or services based on the individually-determined credit score and one or more second terms for at least a second one of the bundle of the plurality of the goods or services, the one or more first terms and the one or more second terms including price and credit terms; and
generate for a user the one or more first terms for the at least first one of the bundle of the plurality of goods or services and the one or more second terms for the at least second one of the bundle of the plurality of goods or services; and
r responsive to capturing and categorizing a new financial transaction made by the user, automatically retraining the machine learning model, recalculating the individually-determined credit score, re-determining the one or more first terms and the one or more second terms.
Claim 29 recites an idea, in part, by:
capturing one or more user transactions for a user at least one payment or transaction tracking method at least one of the user transactions comprising an identification of a purchased good or purchased service, and a price for the purchased good or purchased service, us[ing] a digital profile assigned to the user specifically configured to track informal economic transactions not captured by traditional credit bureau systems;
wherein capturing the one or more user transactions comprises:
receiving,0from a user, an electronic image of a proof of payment for a first one of the user transactions, the first one of the user transactions having no machine-readable record accessible to the one or more processors;
parsing the electronic image using an optical character recognition or intelligent character recognition module and a named entity recognition service to extract a structured transaction record comprising the identification of the purchased good or purchased service and the price for the purchased good or purchased service; and
executing a transaction verification and fraud protection process on the structured transaction record to produce verified user transaction records
storing data associated with the user with reference to the digital profile of the user, including user identification data, the verified user transaction records comprising informal economic activities including at least one of rent payments, utility payments, and bill payments that are not formally recorded in traditional credit bureau systems, user credit score data, user customized pricing data, user recommendations, and user savings data;
processing and categorizing the one or more user transactions into groups including rental payments, bill payments, savings payments and storing the resultant transaction data;
training a machine learning model configured to calculate a credit score associated with the user as a function of the data associated with the user and storing the credit score wherein training machine learning model comprises:
initializing a neural network comprising at least one differential equation and a plurality of neurons having a plurality of independent variables derived from the verified user transaction records, each standardized to a mean of zero and a variance of one, as one or more inputs and one or more weights, the neural network having an input layer comprising a number of input nodes equal to a number of the plurality of independent variables, at least one hidden layer applying a rectifier activation function, and an output neuron applying a sigmoid activation function to produce a continuous-value output representing the credit score;
forward-propagating an observation of a training data set through the neural network to produce a predicted result, wherein a value computed by a first one of the plurality of neurons is input to a second one of the plurality of neurons;
evaluating a gradient descent cost function to determine a value for an error of the neural network by comparing the predicted result to an actual result; and
adjusting values of the one or more weights to reduce the error of the neural network by back-propagating the error via stochastic gradient descent, the one or more weights being updated only after a batch of observations of the training data set has been forward-propagated ,over a plurality of epochs as additional verified user transaction records is acquired;
applying the machine learning model to the verified user transaction records to calculate the credit score of the user;
determining interests for the user in a bundle of a plurality of goods or services based on the one or more user transactions; and
determining one or more first terms for at least a first one of the bundle of the plurality of the goods or services based on the credit score and one or more second terms for at least a second one of the bundle of the plurality of the goods or services, the one or more first terms and the one or more second terms including price and credit terms;
generating for a user the one or more first terms for the at least first one of the bundle of the plurality of goods or services and the one or more second terms for the at least second one of the bundle of the plurality of goods or services; and
responsive to capturing and categorizing a new user transaction made by the user, automatically retraining the machine learning model, recalculating the credit score, re-determining the one or more first terms and the one or more second terms.
And similarly, claim 33 recites an idea, in part by:
capturing user transactions for a user at least one of the user transactions comprising an identification of a purchased good or purchased service, and a price for the purchased good or purchased service, wherein us[ing] a digital profile assigned to the user; and
wherein capturing the user transactions comprises:
receiving, from a user, an electronic image of a proof of payment for a first one of the user transactions, the first one of the user transactions having no machine-readable record accessible to the one or more processors;
parsing the electronic image using an optical character recognition or intelligent character recognition module and a named entity recognition service to extract a structured transaction record comprising the identification of the purchased good or purchased service and the price for the purchased good or purchased service; and
executing a transaction verification and fraud protection process on the structured transaction record to produce verified user transaction records;
storing data associated with the user with reference to the digital profile of the user, including user identification data, the verified user transaction records comprising informal economic activities including at least one of rent payments, utility payments, and bill payments that are not formally recorded in traditional credit bureau systems, user credit score data, user customized pricing data, user recommendations, and user savings data;
processing and categorizing the user transactions into groups including one or more of an extant bundled financial product opted into by the user, mortgage repayments, pension contribution payments, insurance premium payments, rental payments, bill payments, savings payments and storing the user transactions into the groups;
training a machine learning model configured to calculate a credit score associated with the user as a function of the user transactions and data associated with the user and storing the credit score wherein the training machine learning mode comprises:
initializing a neural network comprising at least one differential equation and a plurality of neurons having a plurality of independent variables derived from the verified user transaction records, each standardized to a mean of zero and a variance of one, as one or more inputs, and one or more weights, the neural network having an input layer comprising a number of input nodes equal to a number of the plurality of independent variables, at least one hidden layer applying a rectifier activation function, and an output neuron applying a sigmoid activation function to produce a continuous-value output representing the credit score;
forward-propagating an observation of a training data set through the neural network to produce a predicted result, wherein a value computed by a first one of the plurality of neurons is input to a second one of the plurality of neurons;
evaluating a gradient descent cost function to determine a value for an error of the neural network by comparing the predicted result to an actual result; and
adjusting values of the one or more weights to reduce the error of the neural network by back-propagating the error via stochastic gradient descent, the one or more weights being updated only after a batch of observations of the training data set has been forward-propagated, over a plurality of epochs as additional verified user transaction records is acquired;
applying the machine learning model to the verified user transaction records to calculate the credit score of the user;
determining interests for the user in a bundle of a plurality of goods or services based on the user transactions; and
determining one or more first terms for at least a first one of the bundle of the plurality of the goods or services based on the credit score and one or more second terms for at least a second one of the bundle of the plurality of the goods or services, the one or more first terms and the one or more second terms including price and credit terms;
generating for a user the one or more first terms for the at least first one of the bundle of the plurality of goods or services and the one or more second terms for the at least second one of the bundle of the plurality of goods or services; and
responsive to capturing and categorizing a new user transaction made by the user, automatically retraining the machine learning model, recalculating the credit score, re-determining the one or more first terms and the one or more second terms.
The steps recited above under Step 2A prong 1 of the analysis under the broadest reasonable interpretation covers commercial or legal interactions (including advertising, marketing or sales activities or behaviors; business relations) for calculating credit scores for an individual and offering one or more terms for goods or services based on the information associated with the user, but for the recitation of generic computer components. Other than reciting generic computer components and a machine learning model/unit nothing in the claim elements are directed towards anything other than commercial or legal interactions. If a claim limitation, under its broadest reasonable interpretation covers commercial or legal interactions, then it falls within the “Certain Methods of Organizing Human Activities” groupings of abstract ideas. Accordingly, the claims recite an abstract idea.
Step 2A – Prong Two (Does the claim recite additional elements that integrate the judicial exception into a practical application?) - This judicial exception is not integrated into a practical application. In particular, the claims only recite the additional elements of a memory unit, a computer-readable storage media, one or more processers, processing module, an API gateway comprising a load balancer and service register, display for a graphical user interface, and machine learning model/unit. The memory unit, computer-readable storage media, one or more processers, processing module, API gateway, display for a graphical user interface, and machine learning model/unit are recited at a high level of generality such that it amounts to no more than mere instructions to apply the exception using generic computer components and limits the judicial exception to the particular environment of computers. With respect to the plurality of containerized microservices, API gateway comprising the load balancer and service register, As MPEP 2106.05(f) Use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application or provide significantly more. See Affinity Labs v. DirecTV, 838 F.3d 1253, 1262, 120 USPQ2d 1201, 1207 (Fed. Cir. 2016) (cellular telephone); as is the case of the instant application in which generic computer components are being added after the fact. Further, mere instructions to apply the judicial exception using generic computer components and limiting the judicial exception to a particular environment are not indicative of a practical application (see MPEP 20106.05(f) and MPEP 20106.05(h)). The specification does not provide any indication that the memory unit, computer-readable storage media, one or more processers, processing module, API gateway, display for a graphical user interface, and machine learning model/unit is other than generic computer components as described in pages 23 and 25-26 as an example. Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claims are directed towards an abstract idea.
Step 2B (Does the claim recite additional elements that amount to significantly more than the judicial exception?) - The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because, as discussed above, with respect to integration of the abstract idea into a practical application, using the memory unit, computer-readable storage media, one or more processers, processing module, API gateway, display for a graphical user interface, and machine learning model/unit perform the steps recited above under Step 2A Prong 1 of the analysis amounts to no more than mere instructions to apply the exception using generic computer components and limits the idea to the computer environment. Mere instructions to apply an exception using a generic computer components and limiting an idea to a particular environment does not provide an inventive concept. The additional elements have been considered separately, and as an ordered combination, and do not add significantly more (also known as an “inventive concept”) to the judicial exception. The training and use of the machine learning and defining a neural network is merely using the computer and model to perform repetitive calculations and analyze data akin Performing repetitive calculations, Flook, 437 U.S. at 594, 198 USPQ2d at 199 (recomputing or readjusting alarm limit values);, further, MPEP 2106.05(d)(ii) provides that receiving and transmitting data over a network (see buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network), and Ultramercial, 772 F.3d at 716, 112 USPQ2d at 1755 (updating an activity log); are well-understood routine and conventional, similar to the instant application claims which recites and sending and receiving data over network, and storing and retrieving information from the memory unit for calculating a credit score. Further, the displaying step falls to transform the claims into patent eligible material, as this is part of the field of use and technical environment in which the abstract idea is being implement and does not result in an improvement to additional elements (see MPEP 2106.05(h) Electric Power Group court decision). Thus, the claims are not patent eligible.
The dependent claims have been given the full analysis including analyzing the additional limitations both individually and in combination as a whole. For instance, claims 2-9, 11-14, 30-32, and 34-36 further define the abstract idea and environment in which the idea is being limited to and are all steps that fall within the “Certain Methods of Organizing Human Activities” groupings of abstract ideas similar to above. Claim 10 recites training the machine learning model at high level of generality such it amounts to using a computer as tool to perform repetitive calculations and analyze data, similar to as discussed above. Claims 16-28 are substantially similar to claims 2-14, and ineligible for the same reasons. The Dependent claims when analyzed both individually and in combination are also held to be patent ineligible under 35 U.S.C. 101 for the same reasoning as above and the additional recited limitations fail to establish that the claims are not directed to an abstract idea. The additional limitations of the dependent claims when considered individually and as an ordered combination do not amount to significantly more than the abstract idea.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GREGORY S CUNNINGHAM II whose telephone number is (313)446-6564. The examiner can normally be reached Mon-Fri 8:30am-4pm.
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.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Bennett Sigmond can be reached at 303-297-4411. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
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.
GREGORY S. CUNNINGHAM II
Primary Examiner
Art Unit 3694
/GREGORY S CUNNINGHAM II/Primary Examiner, Art Unit 3694
Prosecution Insights