Abstract: A method including inputting, into a phrase recognition model comprising a neural network, a vector comprising a plurality of ngrams of text. The method also includes applying, using the phrase recognition model, a filter to the plurality of ngrams during execution. The filter has a skip word setting of at least one. The method also includes determining, based on the skip word setting, at least one ngram in the vector to be skipped to form at least one skip word. The method also includes outputting an intermediate score for a set of ngrams that match the filter. The method also includes calculating a scalar number representing a semantic meaning of the at least one skip word. The method also includes generating based on the scalar number and the intermediate score, a final score for the set of ngrams. A computer action is performed using the final score.

Abstract: A system and method allows an offeror to identify parameters for determining whether a user could benefit from an offer, and transaction data for each of several users, and the parameters, are used to determine which of several users could benefit from the offer. The offer is presented to those users, along with the data of how the user could benefit, and the system and method allows the user to accept the offer, transmitting information to the offeror or his agent if the user accepts the offer. Payment may be made based on the offers, acceptances, or both.

Abstract: A method including receiving, from a server, a backoff data packet including first information and second information. The first information includes a backoff command to cease transmitting at least some requests to a service executing on the server, and The second information includes metadata. The method also includes adding the backoff data packet to cached data stored in a cache. The method also includes receiving, from a client, a request data packet including a request for the service. The method also includes comparing request information, associated with the request data packet, to the cached data, including at least comparing the request information to the metadata. The method also includes blocking, responsive to a match between the request information and the cached data, the request data packet from being transmitted to the service.

Abstract: The one or more embodiments provide for a method, system, and computer program product, an intent, generated by a large language model from a text, is received from a user device as a first input to an advice planner. A state of an account is received as a second input to the advice planner. The advice planner classifieds the intent into a domain corresponding to the intent, and generates, as output, a plan comprising a first set of action logic associated with the domain. Each action logic is a discrete step in an ordered sequence for achieving a desired state of the account. The advice planner forwards the plan to the large language model (LLM). The large language model receives the plan as input and generates advice in a natural language format as output. The advice is then forwarded to the user device.

Abstract: A method implements a table discovery service. The method includes receiving a query string, converting the query string to a query graph, and identifying a selected graph, of a set of graphs, that matches the query graph. The method further includes transmitting a notification identifying a previously generated table corresponding to the selected graph, receiving a notification response to accept the previously generated table, and transmitting data from the previously generated table in response to the query string.

Abstract: Systems and methods are disclosed for managing a generative artificial intelligence (AI) model. Managing the generative AI model may include training or tuning the generative AI model before use or managing the operation of the generative AI model during use. Training or tuning a generative AI model typically requires manual review of outputs from the model based on the queries provided to the model to reduce hallucinations generated by the generative AI model. Once the model is in use, though, hallucinations still occur. Use of a confidence (whose generation is described herein) to train or tune the generative AI model and/or manage operation of the model reduces hallucinations, and thus improves performance, of the generative AI model.

Abstract: Systems and methods are disclosed for tuning a generative artificial intelligence (AI) model based on a knowledge base. Instead of manually generating questions relevant to the knowledge base, providing those questions to the generative AI model, and manually reviewing the answers generated by the generative AI model in order to tune the generative AI model over many iterations, a natural language processing model may be configured to leverage the knowledge base to automatically generate questions and answers based on the knowledge base. In this manner, the natural language processing model is able to generate tuning data that may be used to automatically tune the generative AI model. The systems and methods also disclose automatic tuning of the generative AI model, including testing and feedback that may be used to improve tuning of the generative AI model.

Abstract: A content editor for generating content including root blocks and nested blocks is disclosed. The content editor can generate a deployment that includes the content. The content editor can generate user interface code configured to edit the content. The content editor can receive updates to the content and update the root blocks and nested blocks. The updated root blocks and nested blocks can be used to generate updated content for editing and/or can be deployed to end-users.

Abstract: Systems and methods that may be used to provide multitenant key derivation and management using a unique protocol in which key derivation may be executed between the server that holds the root key and a client that holds the derivation data and obtains an encryption key. In one or more embodiments, the derivation data may be hashed. The disclosed protocol ensures that the server does not get access to or learn anything about the client's derived key, while the client does not get access to or learn anything about the server's root key.

Abstract: A tax data collection system includes a navigation module configured to obtain user data. The system also includes a data graph including information relating to the user data. The system further includes a knowledge engine configured to map the user data onto a data model using the information from the data graph. Moreover, the system includes an inference engine configured to suggest a system action by analyzing at least the data model after the user data has been mapped thereon.

Abstract: A system receives a request for payment of a transaction between a vendor and a consumer, and sends a first request to a database associated with the online service for historical transactions and personal attributes of the vendor concurrently with sending a second request to a number of third-party services for credit information and personal attributes of the consumer. The system receives information responsive to the first and second requests from the database and the third-party services, respectively, and obtains a risk score for the transaction based on an application of one or more risk assessment rules to the received information by a machine learning model trained with at least the historical transactions and the personal attributes of the vendor. In some aspects, the system determines whether to advance funds to the vendor, prior to requesting payment from a consumer account, based at least in part on the risk score.

Abstract: Certain aspects of the present disclosure provide techniques for handling crash events in a software application using application-agnostic machine learning models. An example method generally includes receiving a data set of crash reports from a software application for analysis. Using a first neural network, a representation of each respective crash report in the data set is generated. The data set of crash reports and a mapping between functions in the software application and a multidimensional space are input into the first neural network. Each respective crash report in the data set is classified using a second neural network and the representation of each crash report in the data set. One or more actions are taken with respect to the software application based on the classifying each respective crash report in the data set.

Abstract: A method includes training, using first real data objects, a generative adversarial network having a generator model and a discriminator model to create a trained generator model that generates realistic data, and training, using adversarial data objects and second real data objects, the discriminator model to output an authenticity binary class for the adversarial data objects and the second real data objects. The method further includes deploying the discriminator model to a production system. In the production system, the discriminator model outputs the authenticity binary class to a system classifier model.

Abstract: Methods, systems and articles of manufacture for efficiently calculating an electronic tax return, such as within a tax return preparation system. A computerized tax return preparation system accesses taxpayer-specific tax data from a shared data store. The system executes a tax calculation engine configured to perform a plurality of tax calculations based on a tax calculation graph and the taxpayer-specific tax data from the shared data store. The system is configured to perform only the calculations in the tax calculation graph which are changed by new taxpayer-specific tax data received since the preceding tax calculation executed by the tax calculation engine. The system may also determine whether the new taxpayer-specific tax data does, or does not change the calculated tax return and the reason why.

Abstract: A processor may receive user interaction data of a user for a plurality of electronically-presented offers. The processor may generate a plurality of labels, the generating comprising generating a label for each respective offer according to a comparison of the quality of the user interactions of the respective offer to the frequency of the user interactions of the respective offer. Each label may be a positive label or a negative label. The processor may determine whether the generating produced both positive and negative labels. The processor may select one of a plurality of available ML models, wherein a two-class ML model is chosen in response to determining that the generating produced both positive and negative labels and a one-class ML model is chosen in response to determining that the generating did not produce both positive and negative labels. The selected ML model may be trained and/or may be used to process user profile data and provide recommendations.

Abstract: Certain aspects of the disclosure provide systems and methods for training an information extraction transformer model architecture directed to pre-training a first multimodal transformer model on an unlabeled dataset, training a second multimodal transformer model on a first labeled dataset to perform a key information extraction task processing the unlabeled dataset with the second multimodal transformer model to generate pseudo-labels for the unlabeled dataset, training the first multimodal transformer model based on a second labeled dataset comprising one or more labels, the pseudo-labels generated, or combinations thereof to generate a third multimodal transformer model, generating updated pseudo-labels based on label completion predictions from the third multimodal transformer model, and training the third multimodal transformer model using a noise-aware loss function and the updated pseudo-labels to generate an updated third multimodal transformer model.

Abstract: Certain aspects of the present disclosure provide techniques for improving a user experience of an application. Embodiments include receiving, from a user and via a user interface, a request for informational content related to a step in a workflow within the application. Embodiments include determining an identifier associated with the step. Embodiments include retrieving a reference document based on the request. Embodiments include accessing metadata associated with the reference document to identify context information associated with the identifier. Embodiments include displaying a portion of the reference document to the user within the user interface based on the context information, wherein the portion of the reference document comprises the informational content.

Abstract: Techniques for detecting fraud may include obtaining a merchant's financial data; determining, via a machine learning model, a first prediction of the merchant's industry; generating a first probability matrix based on the first prediction and the declared information regarding the merchant's industry; determining, via the machine learning model, a second prediction of the merchant's industry; generating a second probability matrix based on the second prediction and the declared information regarding the merchant's industry; obtaining a declared industry of a subject merchant in a runtime environment; determining, via the machine learning model, a predicted industry for the subject merchant; obtaining, based on the declared industry and the predicted industry of the subject merchant, a first value from the first probability matrix and a second value from the second probability matrix; and labeling the subject merchant for further investigation.

Abstract: Systems and methods here may be used for pre-processing images, including using a computer for receiving a pixelated image of a paper document of an original size, downscaling the received pixelated image, employing a neural network algorithm to the downscaled image to identify four corners of the paper document in the received pixelated image, re-enlarging the downscaled image to the original size, identifying each of four corners of the paper document in the pixelated image, determining a quadrilateral composed of lines that intersect at four angles at the four corners of the paper document in the pixelated image, defining a projective plane of the pixelated image, and determining an inverse transformation of the pixelated image to transform the projective plane quadrilateral into a right angled rectangle.

Abstract: A system deploying a machine learning technique that utilizes known code graph and abstract syntax tree pairs for known JSON objects to learn a function for predicting a corresponding abstract syntax tree from a new JSON object. The predicted abstract syntax tree is used to generate code for formatting the new JSON object into a standardized data structure.