Back to results
US 2026/0289690 A1Published application
System
AssigneeSoftBank Group Corp.
InventorHiroyuki MIZUNO
Life of this patent
Estimating the term…
Abstract
A system includes a processor that is configured to confirm requirements for verifying a financial status of an organization or an enterprise based on performance data of the organization or the enterprise, use a generative AI model to automatically generate audit check items and to preliminarily confirm applicability of the audit check items, and use an emotion engine that recognizes an emotion of a user and provides advice according to the emotion of the user.
Claims20
A system comprising:
- circuitry configured to:
- receive, via a communication interface coupled to a packet-switched network, structured performance data from a terminal device, the structured performance data representing operational conditions of an organization;
- verify a structure and format of the structured performance data, detect missing values and inconsistencies, and convert the structured performance data into an analysis-ready format;
- construct context information for input to a generative neural network model by aggregating features derived from the analysis-ready format, the context information comprising at least a target period indicator, a category identifier, and an anomaly tendency indicator;
- generate, using the generative neural network model and the context information, a plurality of check item candidates in natural language;
- score each check item candidate against condition data stored in a storage device to compute an applicability score, and extract check items for which the applicability score satisfies a threshold;
- transmit the extracted check items to the terminal device via the communication interface;
- receive feedback information from the terminal device comprising user-provided corrections and evaluation data with respect to the extracted check items, and update parameters of the generative neural network model using training data constructed from the feedback information and the context information;
- and
- estimate an emotional state of a user based on input signals received from the terminal device, and generate advice content adapted to the estimated emotional state using the generative neural network model.
7 more elements
A system comprising:
- circuitry configured to:
- receive, via a communication interface coupled to a packet-switched network, structured performance data from a terminal device;
- verify a structure and format of the structured performance data by checking field presence, data type conformance, and value range constraints, detect missing values and flag non-imputable records, and convert validated records into an analysis-ready format by normalizing numerical values and aligning categorical codes using a reference mapping table;
- construct context information by aggregating features from the analysis-ready format into a textual summary comprising a target period indicator, a category identifier, and an anomaly tendency indicator, tokenizing a user-provided prompt string, and concatenating the tokenized prompt string with the textual summary;
- generate, using a transformer-based generative neural network model and the context information, a plurality of check item candidates in natural language by executing forward passes over the context information and decoding output token sequences;
- score each check item candidate against condition data stored in a storage device using a learned text classifier to compute an applicability score, tag each candidate as applicable, partially applicable, or not applicable, and transmit applicable check items to the terminal device;
- receive feedback information from the terminal device comprising user corrections and evaluation data, construct training data pairing the context information with user-approved outputs, and fine-tune the generative neural network model using backpropagation with an adaptive optimization algorithm;
- and
- estimate an emotional state of a user by applying an emotion identification model to input signals from the terminal device, classify the emotional state, and generate advice content adapted to the classified emotional state using the generative neural network model.
6 more elements
A method comprising:
- receiving, via a communication interface coupled to a packet-switched network, structured performance data from a terminal device, the structured performance data representing operational conditions of an organization;
- verifying a structure and format of the structured performance data, detecting missing values and inconsistencies, and converting the structured performance data into an analysis-ready format;
- constructing context information for input to a generative neural network model by aggregating features derived from the analysis-ready format, the context information comprising at least a target period indicator, a category identifier, and an anomaly tendency indicator;
- generating, using the generative neural network model and the context information, a plurality of check item candidates in natural language;
- scoring each check item candidate against condition data stored in a storage device to compute an applicability score, and extracting check items for which the applicability score satisfies a threshold;
- transmitting the extracted check items to the terminal device via the communication interface;
- receiving feedback information from the terminal device comprising user-provided corrections and evaluation data with respect to the extracted check items, and updating parameters of the generative neural network model using training data constructed from the feedback information and the context information;
- and
- estimating an emotional state of a user based on input signals received from the terminal device, and generating advice content adapted to the estimated emotional state using the generative neural network model.
6 more elements
Dependencies are inferred from the claim text.