Abstract: A method may include extracting, from a document, a first key-value pair including a key and a first value and corresponding to a first confidence score, extracting a second key-value pair including the key and a second value corresponding to a second confidence score, classifying a first match probability for the first key-value pair and a second match probability for the second key-value pair, generating a first calibrated confidence score for the first confidence score and a second calibrated confidence score for the second confidence score by transforming, using precision lookup tables constructed from training records, the first match probability to the first calibrated confidence score and the second match probability to second calibrated confidence score, selecting, using the first and second calibrated confidence scores, one of the first key-value pair and the second key-value pair, and presenting, in a graphical user interface (GUI), the selected key-value pair.

Abstract: A method including receiving a context-specific query, specific to the user, from a user device of a user. The method also includes creating a computer-readable data structure for storing data. The computer-readable data structure is specific to the context-specific query. The method also includes determining a feature related to the context-specific query. The method also includes retrieving a user value for the feature. The user value is retrieved from a data repository storing user-specific data that includes the user value. The method also includes modifying the computer-readable data structure to generate a modified computer-readable data structure by adding the feature and the user value for the feature to the computer-readable data structure. The method also includes applying a generative model to the modified computer-readable data structure to generate an output. The output includes a context-specific answer to the context-specific query. The method also includes returning the output.

Abstract: Systems and methods are provided for enhancing vector searches for question-and-answer platforms.

Abstract: Systems and methods are described for batch materialization of an incremental change data capture (CDC) changeset. The primary keys are extracted from the incremental CDC changeset and an indication of the extracted primary keys are broadcast to a plurality of executors. The primary keys may be added to Bloom filter or a plurality of Bloom filters that are broadcast to the executors. Each executor filters a baseline data table based on the extracted primary keys to generate a baseline match dataframe with all primary keys matching the extracted primary keys, and a baseline unmatched dataframe with all primary keys not matching the extracted primary keys. Each executor receives a partitioned incremental CDC changeset and combines the changes with the baseline match dataframe to produce a final changed baseline data table.

Abstract: The disclosed embodiments provide an augmented reality system. The augmented reality system includes a camera that captures a set of images of one or more physical documents. The augmented reality system also includes a processing apparatus that enables use of the augmented reality system by a user in performing financial management associated with data in the one or more physical documents.

Abstract: At least one processor can receive an indication of processing to be performed in a natural language processing (NLP) machine learning (ML) pipeline, determine a first model class and a second model class for the processing. The first model class can be at a first hierarchical level of an ML hierarchy schema, and the second model class can be at a second hierarchical level of the ML hierarchy schema. The at least one processor can prepare a dictionary in a memory in communication with the at least one processor, which can comprise populating the dictionary with all required artifacts of the first model class and a subset of required artifacts of the second model class, wherein the second model class requires at least one of the required artifacts of the first model class. The at least one processor can perform NLP on text using the ML model and the dictionary.

Abstract: At least one processor may receive a plurality of review comments, identify at least a subset of the plurality of review comments representing at least one positive review, and identify at least one product indicated by the at least one positive review. The at least one processor may determine, from data produced by an inventory system, that the at least one product has a low inventory level. In response to the determining, the at least one processor may generate a message configured to automatically order the at least one product from a supplier system and send the message to the supplier system, thereby automatically ordering the at least one product.

Abstract: A method includes receiving, at a server from a user device, a user query to a large language model (LLM), creating an LLM query from the user query, inserting a system prohibited request into the LLM query to generate a revised LLM query, and sending the revised LLM query to the LLM. The method further includes receiving, from the LLM, a first LLM response to the LLM query, testing the first LLM response to detect whether a prohibited response to the system prohibited request is included in the first LLM response, and setting a prompt injection signal based on whether the prohibited response to the system prohibited request is included in the first LLM response.

Abstract: Systems and methods may be used to generate and use a structured form representation and structured metadata. The structured form representation and structured metadata may include information relevant to a particular context and may be used to update document templates, import new documents and update document versions into software, automate data entry for document completion, update records to include new and or updated information, and provide other functionality of an information service.

Abstract: A method for rule-based composition of user interfaces. A machine-learned rule repository is established based on previously observed combinations of UI states, UI features, and user features. A classifier classifies users into segments. Each segment includes users for which a combination of user features and UI states are defined. A first machine learning model (MLM) estimates a user segment-content preference including preferred UI content. A second MLM estimates a seen content-seen content similarity UI content preferences estimated according to prior UI content a user has seen. Based on the UI state and based on the user ID, rule-based recipes are obtained. Each rule-based recipe specifies a corresponding UI content suitable for an interaction between the user and the interface. A selected rule-based recipe is selected from the rule-based recipes. Specific UI content specified by the selected rule-based recipe is obtained, and the interface is updated with the specific UI content.

Abstract: A method that includes extracting image features of a document image, executing an optical character recognition (OCR) engine on the document image to obtain OCR output, and extracting OCR features from the OCR output. The method further includes executing an anomaly detection model using features including the OCR features and the image features to generate anomaly score, and presenting anomaly score.

Abstract: Systems and methods that generate an enhanced search report, which provides a significant improvement over conventional searches are provided. The systems and methods leverage a combination of keyword-based searches using search engine APIs to return URLs that are hit by the searches, large language models to generate additional keywords based on contextual information, and web crawlers to search the returned URLs with the additional keywords. The enhanced search report is not confined to the conventional keyword-URL match, but instead provides a sophisticated capture and compilation of additional useful information.

Abstract: A sequence of data entry screens are configured to collect the data from a user. The method and system receive data entered by a user into a data entry screen. The method and system then determine metrics of the collected data, and ranks the collected data and the data entry screens based on the determined metrics. The ranking is then used to display the next best screen in the sequence for collecting data.

Abstract: A processor may receive a request to display a user interface (UI) for a user account. The processor may determine a context for the UI from information specific to the user account. The processor may select one of a plurality of UI elements based on the determining. In some embodiments, the selecting may include evaluating a value function taking the information specific to the user account and information about the context as inputs. The processor may cause the UI including the selected one of the plurality of UI elements to be displayed in response to the request.

Abstract: A method including receiving, at an embedded browser embedded in an application, a request to access data designated by a uniform resource locator (URL) specified by the request. The method also includes intercepting, by a method interceptor, an application programming interface (API) call to access the data designated by the URL. Intercepting is performed prior to execution of the API call. The API call is performable by an API of the embedded browser. The method also includes comparing, by the method interceptor, a domain specified by the URL to an list of allowed domains. The method also includes blocking, by the method interceptor and responsive to the domain failing to be a member of the list of allowed domains, the API call. Blocking is performed by the method interceptor preventing the API call from passing to the API.

Abstract: Aspects of the present disclosure relate to detecting spam messages. Embodiments include creating condensed vector representations of messages; calculating similarity scores for each message relative to other messages using the vector representations; associating messages of the plurality of messages with a grouping based on the calculated similarity score for the messages within the grouping exceeding a threshold; determining that a grouping of messages are spam messages by comparing an amount of messages of the grouping of messages sent within a first time window to an amount of messages of the grouping of messages sent within a second time window, wherein the second time window comprises a time period preceding the first time window; providing the identified spam message to a machine learning model; and training the machine learning model by iteratively adjusting parameters of the model based on tracking the identified spam message through multiple layers of the model.

Abstract: A method including applying a propensity model to a subject vector to generate a propensity value estimating a probability that a subject will perform an action. The subject vector has a data structure having features storing information regarding the subject. The method also includes applying a Shapley additive explanation tool to the propensity model to generate a subset of the features that contributed to the propensity value more than a remaining set of the features. The method also includes selecting an actionable feature from the subset of the features. The actionable feature includes a feature in the subset that an entity is able to influence. The method also includes applying a correlation model to labels for training data and the actionable feature to generate an output that describes a reason why the subject performs the action. The method also includes presenting the actionable feature and the output.

Abstract: An API security system that implements techniques for learning sensitive information fields from a group of API requests. The API security system implementing security measures for protecting the identified sensitive information fields.