Method and system for generating executive summaries and data visualizations for annual product quality reports
Inventors
Singh, Ankit • Subber, Waad • Ghule, Niranjan • Kelkar, Shraddha
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12639345-B2
Publication Date
2026-05-26
Expiration Date
Abstract
The present disclosure provides a system and method for generating text summary and data visualizations for executive summaries of annual product quality reports based on an input query from a user. The method for generating text summary for annual product quality reports (APQR) of an enterprise, comprising: providing one or more indexes corresponding to one or more categories of data stored in a database, receiving a query from one or more users, wherein the query is a text and based on the at least one of the indexes of the one or more categories of data, extracting one or more relevant textual content from the query using natural language processing, querying the database for retrieving the one or more indexes of the relevant textual content, providing one or more few shot prompts for generating an output in a predetermined format based on the query received from the one or more users, inputting, the one or more few shot prompts and the retrieved one or more indexes to a large language model (LLM), generating a text summary based on the input to the LLM and rendering the generated text summary at a user device.
Core Innovation
The invention relates to generating text summary for annual product quality reports (APQR) of an enterprise by combining category-based retrieval from a database with natural language processing and a large language model (LLM). One or more indexes corresponding to one or more categories of data stored in a database are provided, and a text query from one or more users is received based on at least one of the indexes of the one or more categories of data. Natural language processing is used to extract one or more relevant textual content from the query, and the database is queried to retrieve one or more indexes of the relevant textual content.
The retrieved indexes and the query are then used to provide one or more few shot prompts for generating an output in a predetermined format. The one or more few shot prompts comprise at least one or more queries received from the one or more users for the past APQR and at least one or more text summaries of the past APQR. The one or more few shot prompts and the retrieved one or more indexes are input to an LLM to generate a text summary.
The generated text summary is rendered at a user device as an executive-style output consistent with past APQR content and formatting. In further embodiments, business intelligence-driven generation of one or more data visualizations is supported in addition to the text summary, using retrieved indexes and the few shot prompts, with rendering at the user device and optional transmission to one or more authorities for regulatory compliance.
Claims Coverage
The document includes three independent claims (method, system, and non-transitory computer-readable storage medium) for generating APQR text summaries using category indexes, NLP-based extraction, retrieval, and few-shot prompting with past APQR queries and text summaries. Across the independent claims, the main inventive features remain consistent, with additional refinements covering optional data visualizations, update behavior tied to retrieved indexes changing, output format/content matching past APQR summaries, vector database retrieval with an adjustable similarity threshold, and transmission to authorities for regulatory compliance.
Category-indexed APQR text summary generation from user query
Providing one or more indexes corresponding to one or more categories of data stored in a database; receiving a query from one or more users, wherein the query is a text and based on the at least one of the indexes of the one or more categories of data; extracting one or more relevant textual content from the query using natural language processing; querying the database for retrieving the one or more indexes of the relevant textual content.
Few-shot prompting using past APQR queries and past APQR text summaries in predetermined format
Providing one or more few shot prompts for generating an output in a predetermined format based on the query received from the one or more users, wherein the one or more few shot prompts comprises at least one or more queries received from the one or more users for the past APQR and at least one or more text summaries of the past APQR; inputting, the one or more few shot prompts and the retrieved one or more indexes to a large language model (LLM); generating a text summary based on the input to the LLM.
Rendering an LLM-generated APQR executive text summary at a user device
Rendering the generated text summary at a user device.
Business intelligence-driven data visualization generation with APQR retrieval prompts
Generating one or more data visualizations via a business intelligence service from the retrieved one or more indexes and the few shot prompts and rendering the one or more data visualizations on the users’ devices.
Regeneration of text summary and data visualizations when retrieved indexes change
Generating an updated text summary and data visualizations when the retrieved one or more indexes for the one or more relevant textual content change.
Format and content constraint to match past APQR text summaries
Generating text summaries and data visualizations in the same format and content as past APQR text summaries.
Vector database retrieval of relevant past APQR content using similarity threshold
Retrieving one or more whole or partial past APQR documents by comparing query and stored embeddings in a vector database and selecting relevant data when a similarity score based on vector difference exceeds an adjustable threshold.
Transmission of generated APQR text summary and data visualizations to authorities for regulatory compliance
Transmitting a generated text summary and one or more data visualizations to one or more authorities for regulatory compliance.
Across the independent claims, the inventive coverage centers on category-based indexing and NLP extraction from a user text query, retrieval of corresponding indexes, and LLM generation using few-shot prompts constructed from past APQR user queries and past APQR text summaries to produce a predetermined-format output rendered at a user device. Dependent claim refinements further broaden coverage to business intelligence-driven data visualizations, conditional regeneration when retrieved indexes change, constraints to match past APQR format and content, vector database retrieval with an adjustable similarity-threshold condition, and transmitting generated outputs to authorities for regulatory compliance.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?