# US 12,602,385 B2 · Context-aware relevance modeling in conversational systems > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,602,385 B2 - Document ID: US12602385B2 - Status: Granted - Kind: B2 - Application number: 18/453,127 - Filed: 2023-08-21 - Published: 2025-02-27 - Granted: 2026-04-14 - Earliest claimed: 2023-08-21 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Hui Wan; Xiaodong Cui; Songtao Lu; Marina Danilevsky Hailpern - Web page: https://patentgrep.com/patent/US12602385B2 ## CPC - G06F 16/24575: {using context} - G06F 16/3329: Natural language query formulation - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A method, computer system, and a computer program product are provided for a context-aware relevancy modelling in conversational systems. A user query is received. A latent static content d is selected from a corpus of content D. A latent set of context C from a set of external context Cu is also selected. A result is generated using a scoring function and using the latent static content d from a corpus D and the latent set of context C from the set of external contexts CU so as to provide a most relevant context-base search response to said user query q. The result provides a most relevant context-base search response to said user query q. A response is then generated based on said result using said scoring function result to said user query q. ## Claims ### Claim 1 A method for providing context-aware relevancy modelling in conversational systems, comprising: receiving a user query q; processing said user query q using a scoring function, wherein said processing include collecting information relating to said user query q from one or more other users; selecting a latent static content d from a corpus of content D and a latent set of context C from a set of external context Cu; generating a result using said scoring function having said latent static content d from said corpus of content D and said latent set of context C from a set of external contexts CU, wherein said result generated is responsive to a most relevant context-base search to said user query q; performing a search using said result; generating a response based on the search performed using said result and using said scoring function, in response to said user query q; and formulating from a dialogue extracted from user previous interactions, selecting said latent static content d from the search performed using said corpus of content D. ### Claim 2 (depends on 1) The method of claim 1, wherein the scoring function has a fusion function and a relevance function component. ### Claim 3 (depends on 1) The method of claim 1, wherein said most relevant context-base search includes an inference component. ### Claim 4 (depends on 1) The method of claim 1, wherein an inference is generated via at least one of a greedy search and beam search. ### Claim 5 (depends on 4) The method of claim 4, wherein said beam search is in a direction from q to d to C and a direction from q to C to d, wherein q is the user query and C is latent set of context and d is a latent static content. ### Claim 6 (depends on 1) The method of claim 1, wherein said response is stored. ### Claim 7 (depends on 6) The method of claim 6, wherein said response that is stored is used to train an artificial intelligence (AI) engine having one or more machine language models. ### Claim 8 (depends on 7) The method of claim 7, wherein one or more machine learning models includes a large language model (LLM). ### Claim 9 (depends on 1) The method of claim 1, given said corpus of content D, and said set of latent context C, and said user query q, wherein said latent set of context C={c1, . . . ,cn} from Cu, that that are the most relevant to the query q. ### Claim 10 A computer system for restoring an interrupted communication session, comprising: one or more processors, one or more computer-readable memories and one or more computer-readable storage media; program instructions, stored on at least one of one or more of a plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to receive a user query q; program instructions, stored on at least one of one or more of a plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to process said user query q using a scoring function, wherein said processing include collecting information relating to said user query q from one or more other users; program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to select a latent static content d from a corpus of content D and a latent set of context C from a set of external context Cu; program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to generate a result using said scoring function having said latent static content d from said corpus of content D and said latent set of context C from a set of external contexts CU, wherein said result generated is responsive to a most relevant context-base search to said user query q; program instructions, stored on at least one of one or more of a plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to perform a search using said result; program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to generate a response based on the search performed using said result and using said scoring function, in response to said user query q; and program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to formulate from a dialogue extracted from user previous interactions, selecting said latent static content d from the search performed using said corpus of content D. ### Claim 11 (depends on 10) The computer system of claim 10, wherein the scoring function has a fusion function and a relevance function component. ### Claim 12 (depends on 10) The computer system of claim 10, wherein a search result that is updated includes an inference component. ### Claim 13 (depends on 10) The computer system of claim 10, wherein the most relevant context-base search includes an inference component. ### Claim 14 (depends on 13) The computer system of claim 13, wherein said search is in a direction q to d to C and a direction from q to C to d wherein q is the user query and C is latent set of context and d is a latent static content. ### Claim 15 (depends on 10) The computer system of claim 10, wherein said response is stored. ### Claim 16 (depends on 15) The computer system of claim 15, wherein a stored result is used to train an artificial intelligence (AI) engine having one or more machine language models. ### Claim 17 (depends on 15) The computer system of claim 15, wherein one or more machine learning models includes a large language model (LLM). ### Claim 18 A computer program product for providing cleansing steps for using a plurality of different transformation assets, the computer program product comprising: one or more non-transitory computer readable storage media, one or more processors and one or more memories; program instructions, stored on at least one of one or more of a plurality of storage media for execution by at least one of the one or more processors via at least one of one or more computer-readable memories, to receive a user query q; program instructions, stored on at least one of one or more of a plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to process said user query q using a scoring function, wherein said processing include collecting information relating to said user query q from one or more other users; program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to select a latent static content d from a corpus of content D and a latent set of context C from a set of external context Cu; program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to generate a result using said scoring function having said latent static content d from said corpus of content D and said latent set of context C from a set of external contexts CU, wherein said result generated is responsive to a most relevant context-base search to said user query q; program instructions, stored on at least one of one or more of a plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to perform a search using said result; program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to generate a response based on the search performed using said result and using said scoring function, in response to said user query q; and program instructions, stored on at least one of the one or more of said plurality of storage media for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, to formulate from a dialogue extracted from user previous interactions, selecting said latent static content d from the search performed using said corpus of content D. ### Claim 19 (depends on 18) The computer program product of claim 18, wherein said scoring function has a fusion function and a relevance function component. ### Claim 20 (depends on 18) The computer program product of claim 18, wherein a first search result that is updated includes an inference component and said inference component is generated via at least one of a greedy search and beam search.