US 12,393,620 B1
Large language model-based question answering system
Saurabh S. Khanwalkar, San Mateo, CA (US); Vipul Gagrani, San Jose, CA (US); Kazem Jahanbakhsh, Redwood City, CA (US); Mahdi Hajiabadi, Vancouver (CA); Jennifer Louie, Mountain View, CA (US); and Mina Naghshnejad, San Ramon, CA (US)
Assigned to Learneo, Inc., Redwood City, CA (US)
Filed by Learneo, Inc., Redwood City, CA (US)
Filed on Dec. 20, 2024, as Appl. No. 18/990,934.
Application 18/990,934 is a continuation of application No. 18/940,680, filed on Nov. 7, 2024.
Claims priority of provisional application 63/549,107, filed on Feb. 2, 2024.
Claims priority of provisional application 63/597,657, filed on Nov. 9, 2023.
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 16/3329 (2025.01)
CPC G06F 16/3329 (2019.01) 19 Claims
OG exemplary drawing
 
1. A system, comprising:
one or more processors configured to:
receive a question;
identify a type of the question;
construct a prompt to provide to a generative model based on the identified type of the question; and
provide output based on a generative response provided by the generative model, wherein the output further comprises a citation, wherein the citation is based at least in part on one or more contextual passages, and wherein:
the one or more contextual passages are determined at least in part by conducting a search of a passage store using an answer and an explanation to the question, wherein the passage store comprises a vector database including passage embedding vectors, and wherein the search of the vector database is conducted based at least in part on generating of an answer-explanation embedding vector representing text of the answer and the explanation to the question; and
providing the output comprises including, in the output, links to internal documents from which the one or more contextual passages were extracted; and
a memory coupled to the one or more processors and configured to provide the one or more processors with instructions.