US 12,259,915 B2
Natural language database generation and query system
Javed Qadrud-Din, Union City, CA (US); Pablo Arredondo, Palo Alto, CA (US); Walter DeFoor, Rockville, MD (US); and Alan deLevie, Washington, DC (US)
Assigned to Casetext, Inc., Eagan, MN (US)
Filed by Casetext, Inc., San Francisco, CA (US)
Filed on Nov. 20, 2023, as Appl. No. 18/515,014.
Application 18/515,014 is a continuation of application No. 18/329,035, filed on Jun. 5, 2023, granted, now 11,860,914.
Claims priority of provisional application 63/487,181, filed on Feb. 27, 2023.
Prior Publication US 2024/0289363 A1, Aug. 29, 2024
Int. Cl. G06F 16/33 (2019.01); G06F 16/3329 (2025.01); G06F 40/289 (2020.01); G06F 40/205 (2020.01)
CPC G06F 16/3329 (2019.01) [G06F 40/289 (2020.01); G06F 40/205 (2020.01)] 19 Claims
OG exemplary drawing
 
1. A method comprising:
receiving via a communication interface a request to generate novel text, the request including a natural language input text element identify in natural language one or more parameters for generating the novel text;
retrieving a plurality of records from a database system based on the natural language input text element, the plurality of records corresponding with a plurality of document portions, each of the document portions being selected from a respective document of a plurality of documents, wherein retrieving the plurality of records from the database system comprises transmitting to a remote text generation system an input query comprehension prompt including query comprehension text determined based on the request and receiving from the remote text generation system a completed query comprehension prompt including a description of the request generated by a text generation model implemented at the remote text generation system;
determining an original natural language output text portion generated based on the plurality of document portions and the natural language input text element;
determining via a processor whether the original natural language output text portion is factually supported by the plurality of document portions;
determining an updated natural language output text portion upon determining that a subset of the original natural language output text portion is not factually supported by the plurality of document portions; and
transmitting via the communication interface a response message including the updated natural language output text portion.