# US 12,762,165 B2 · Utilizing a large language model to perform a query > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,762,165 B2 - Document ID: US12762165B2 - Status: Granted - Kind: B2 - Application number: 18/652,615 - Filed: 2024-05-01 - Published: 2025-03-06 - Granted: 2026-10-13 - Earliest claimed: 2023-09-06 - Assignees: Tiny Fish Inc. - Applicants: Tiny Fish Inc. - Inventors: Shuhao Zhang; Qi Zhai; Mingyang Li - Web page: https://patentgrep.com/patent/US12762165B2 ## CPC - G06F 16/3344: {using natural language analysis} - G06F 16/3325: {Reformulation based on results of preceding query} - G06F 16/345: {Summarisation for human users} - G06F 16/35: Clustering; Classification - G06F 40/205: Parsing - G06F 40/279: Recognition of textual entities - G06F 40/30: Semantic analysis - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A query is received from a client device. A large language model is prompted to generate a plurality of subtopics on the query and to generate a corresponding plurality of keywords for each of the plurality of subtopics. One or more search engines are utilized to perform a plurality of searches utilizing the plurality of subtopics and the corresponding plurality of keywords received from the large language model. A plurality of responses corresponding to the plurality of subtopics and the corresponding plurality of keywords is received from the one or more search engines. The plurality of responses is evaluated based on the corresponding plurality of keywords. ## Claims ### Claim 1 A method, comprising: receiving from a client device a query; providing to a large language model a prompt to generate a plurality of subtopics on the query and to generate a corresponding plurality of keywords for each of the plurality of subtopics; utilizing one or more search engines to perform a plurality of searches utilizing the plurality of subtopics and the corresponding plurality of keywords received from the large language model; receiving from the one or more search engines a plurality of responses corresponding to the plurality of subtopics and the corresponding plurality of keywords; evaluating the plurality of responses based on the corresponding plurality of keywords; providing to the large language model the received query and a corresponding subset of sentences associated with each of the plurality of subtopics, wherein the corresponding subset of sentences are selected from a subset of the plurality of responses; receiving a query response from the large language model that is generated based on the received query and the corresponding subset of sentences associated with each of the plurality of subtopics; and providing to the client device the query response that includes a plurality of links to sources that were used by the large language model to generate the large language model query response. ### Claim 2 (depends on 1) The method of claim 1, further comprising rephrasing the query. ### Claim 3 (depends on 2) The method of claim 2, wherein the query is rephrased to fix grammatical and/or spelling errors. ### Claim 4 (depends on 2) The method of claim 2, wherein the query is rephrased to present the query in an improved format for the large language model prompt. ### Claim 5 (depends on 1) The method of claim 1, wherein a prompt to generate the plurality of subtopics on the query includes a corresponding number of the subtopics and a corresponding number of the keywords for each of the plurality of subtopics. ### Claim 6 (depends on 1) The method of claim 1, further comprising receiving from the large language model the plurality of subtopics and the corresponding plurality of keywords for each of the plurality of subtopics. ### Claim 7 (depends on 1) The method of claim 1, wherein evaluating the plurality of responses includes counting, for each response included in the plurality of responses, a number of times the corresponding plurality of keywords appears in the plurality of responses. ### Claim 8 (depends on 7) The method of claim 7, wherein the plurality of responses includes a plurality of search result snippets. ### Claim 9 (depends on 7) The method of claim 7, wherein evaluating the plurality of responses includes ranking the plurality of responses. ### Claim 10 (depends on 9) The method of claim 9, wherein the plurality of responses is ranked based on the number of times the corresponding plurality of keywords appears in a corresponding response. ### Claim 11 (depends on 9) The method of claim 9, wherein the plurality of responses is ranked based on a corresponding domain associated with the plurality of responses. ### Claim 12 (depends on 9) The method of claim 9, wherein evaluating the plurality of responses includes generating a corresponding subset of responses for each of the plurality of subtopics based on the plurality of ranked responses. ### Claim 13 (depends on 12) The method of claim 12, wherein a top number or a top percentage of the ranked responses is included in the corresponding subset of responses. ### Claim 14 (depends on 12) The method of claim 12, further comprising parsing text included in pages linked to the corresponding subset of responses. ### Claim 15 (depends on 14) The method of claim 14, further comprising ranking sentences included in the parsed texted included in the pages linked to the corresponding subset of responses based on the number of times the corresponding plurality of keywords appears in the sentences. ### Claim 16 (depends on 15) The method of claim 15, wherein each of the plurality of subtopics is associated with a corresponding subset of sentences that are selected from the ranked sentences. ### Claim 17 (depends on 16) The method of claim 16, wherein each of the plurality of subtopics is associated with a word limit. ### Claim 18 (depends on 1) The method of claim 1, further comprising utilizing one or more web agents to parse text associated with one or more online sources. ### Claim 19 (depends on 18) The method of claim 18, wherein at least one of the one or more web agents utilizes login credentials associated with a user of the client device to access the one of the one or more online sources. ### Claim 20 (depends on 1) The method of claim 1, further comprising: receiving from the client device one or more subsequent queries related to the query; and storing the query and the one or more subsequent queries related to the query as a branched query topic. ### Claim 21 (depends on 20) The method of claim 20, further comprising providing access to the branched query topic via a query topic board. ### Claim 22 A system, comprising: a processor configured to: receive from a client device a query; provide to a large language model a prompt to generate a plurality of subtopics on the query and to generate a corresponding plurality of keywords for each of the plurality of subtopics; utilize one or more search engines to perform a plurality of searches utilizing the plurality of subtopics and the corresponding plurality of keywords received from the large language model; receive from the one or more search engines a plurality of responses corresponding to the plurality of subtopics and the corresponding plurality of keywords; evaluate the plurality of responses based on the corresponding plurality of keywords; provide to the large language model the received query and a corresponding subset of sentences associated with each of the plurality of subtopics, wherein the corresponding subset of sentences are selected from a subset of the plurality of responses; receive a query response from the large language model that is generated based on the received query and the corresponding subset of sentences associated with each of the plurality of subtopics; and provide to the client device the query response that includes a plurality of links to sources that were used by the large language model to generate the large language model query response; and a memory coupled to the processor and configured to provide the processor with instructions. ### Claim 23 A computer program product embodied in a non-transitory computer readable medium and comprising computer instructions for: receiving from a client device a query; providing to a large language model a prompt to generate a plurality of subtopics on the query and to generate a corresponding plurality of keywords for each of the plurality of subtopics; utilizing one or more search engines to perform a plurality of searches utilizing the plurality of subtopics and the corresponding plurality of keywords received from the large language model; receiving from the one or more search engines a plurality of responses corresponding to the plurality of subtopics and the corresponding plurality of keywords; evaluating the plurality of responses based on the corresponding plurality of keywords; providing to the large language model the received query and a corresponding subset of sentences associated with each of the plurality of subtopics, wherein the corresponding subset of sentences are selected from a subset of the plurality of responses; receiving a query response from the large language model that is generated based on the received query and the corresponding subset of sentences associated with each of the plurality of subtopics; and providing to the client device the query response that includes a plurality of links to sources that were used by the large language model to generate the large language model query response.