US 12,259,938 B2
Delivering responses to queries expressed in natural language based on a dynamic document corpus
Luke Thomas DeVos, Pittsburgh, PA (US); Timothy James Ireland, II, Bellevue, WA (US); Abigail Lynn Ireland, Bellevue, WA (US); Siu Tang Leung, Seattle, WA (US); Andrei Modoran, Bellevue, WA (US); Brad Steven Ostercamp, Carnation, WA (US); and Jagdeesh Prakasam, Mercer Island, WA (US)
Assigned to Qdeck Inc., Bellevue, WA (US)
Filed by Qdeck Inc., Bellevue, WA (US)
Filed on May 3, 2024, as Appl. No. 18/655,024.
Claims priority of provisional application 63/499,945, filed on May 3, 2023.
Prior Publication US 2024/0370517 A1, Nov. 7, 2024
Int. Cl. G06F 16/00 (2019.01); G06F 16/955 (2019.01); G06F 16/957 (2019.01); G06F 40/295 (2020.01)
CPC G06F 16/9558 (2019.01) [G06F 16/957 (2019.01); G06F 40/295 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for generating responses to user queries with direct citations to primary sources, the method comprising:
accessing a set of documents from a plurality of primary sources via a data feed transmission, wherein data from each source of the plurality of primary sources is transmitted in a different data structure;
determining values for a set of parameters based on data of each document of the set of documents, wherein the set of parameters includes a combination of (1) a plurality of segments of each document, (2) one or more entities identified from the data, and (3) scores characterizing the data;
organizing values of the set of parameters into a searchable data structure;
responsive to receiving a user query at a user interface, inputting a prompt based on the user query into a large-language model;
obtaining an output response as a result of inputting the prompt a the large-language model;
processing the output response in combination with the searchable data structure to identify references to relevant documents from the set of documents; and
embedding, into the output response generated by the large-language model, citations to the references from the plurality of primary sources.