| CPC G06F 16/3349 (2019.01) | 20 Claims |
|
1. A system for improving user interaction with a large language model (LLM) to manage corporate security comprising:
an AI engine comprising:
memory storing the LLM;
computer circuitry configured to:
interact with users by responding to user queries, wherein responding to each of the user queries comprises:
receiving a query from a user;
processing the received query with the LLM to generate an answer to the query;
providing the generated answer to the user;
receiving feedback from the user concerning the provided answer, wherein the feedback identifies one or more quality aspects of the answer including at least one of accuracy, relevance, completeness, and clarity; and
outputting the received feedback and the query;
an enhancement engine including processor circuitry configured to:
receive the query and the feedback output by the AI engine;
generate from the received query and the received feedback at least one if/then pair, wherein:
each of the generated if/then pairs includes an if statement and a then statement; and
each of the generated if/then pairs is generated by:
extracting as the if statement a condition or scenario from the query; and
generating from the feedback as the then statement a response indicating improvements or corrections for guiding answer generation by the AI engine; and
store each of the generated at least one if/then pairs in a database;
wherein the computer circuitry of the AI engine is further configured to:
generate an enhanced answer by:
receiving an additional query from the user;
retrieving from the database, as related if/then pairs, the stored if/then pairs having an if statement related to the received additional query;
combining as an enhanced query the additional query with the then statements of the retrieved if/then pairs;
processing the enhanced query with the LLM to generate an enhanced answer; and
providing the generated enhanced answer to the user.
|