US 12,436,957 B1
Context-specific query response platform using large language models
Samrendra Kumar, Perumbakkam (IN); Karthik Mohan, Scarborough (CA); Frenchy George Thomas, Chennai (IN); Rohit Khanna, Cranbury, NJ (US); Leslie Dias, West Windsor, NJ (US); and Winnie Chacko, Dix Hills, NY (US)
Assigned to Bank of America Corporation, Charlotte, NC (US)
Filed by Bank of America Corporation, Charlotte, NC (US)
Filed on Jul. 26, 2024, as Appl. No. 18/785,964.
Int. Cl. G06F 16/242 (2019.01); G06F 16/245 (2019.01); G06F 16/2453 (2019.01); G06F 16/2457 (2019.01)
CPC G06F 16/24575 (2019.01) [G06F 16/2425 (2019.01); G06F 16/24539 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A computing platform comprising:
at least one processor;
a communication interface communicatively coupled to the at least one processor; and
a memory storing computer-readable instructions that, when executed by the at least one processor, configure the computing platform to:
train, based on historical query information, a query response model, wherein training the query response model configures the query response model to output query responses based on input of query prompts;
receive, from a user device, a user query;
generate, based on one or more historical embeddings, an embedding corresponding to the user query;
generate a query prompt corresponding to the user query, wherein the query prompt comprises:
the embedding corresponding to the user query; and
a plurality of embeddings corresponding to a subset of the historical query information;
execute, using as inputs the query prompts, the query response model to output a query response;
send, to the user device, the query response, wherein sending the query response to the user device causes display of a user interface including the query response on a display of the user device;
identify, based on sending the query response, a status of the query response indicating whether the query response satisfies the user query, and in response:
based on identifying that the query response satisfies the user query, store, to a long-term memory component, the query response, or
based on identifying that the query response does not satisfy the user query, output an alternative query response; and
update, based on the query response and the status of the query response, the query response model.