Abstract
A method includes: chunking each resource into chunks; inputting the chunks into a model; for each chunk, generating an embedding vector; receiving a query message from a user device; inputting the query message into the model; generating a query embedding vector for the query message; executing a semantic search on the query embedding vector by: inputting the query embedding vector and the embedding vector for each chunk to the model; for each chunk, generating a chunk similarity score; determining a subset of chunks based on the chunk similarity score for each chunk of the resource; inputting the subset of chunks of a subset of resources to a second model; generating a query response message based on the subset of chunks of the resource; and communicating the query response message to the user device.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
KUO, JIA YI; HUANG, JIANYANG; GOPALAKRISHNA, KARTHIK; GANDHI, BHUVANJEET SINGH; GUHA, PAROMITA; CHANDEL, ASHUTOSH KUMAR; AGRAWAL, ARPITA; and SAXENA, RITU, "METHOD, SYSTEM, AND COMPUTER PROGRAM PRODUCT FOR LARGE LANGUAGE MODEL (LLM)-ENABLED SEARCHING", Technical Disclosure Commons, (July 28, 2025)
https://www.tdcommons.org/dpubs_series/8399