# US 11,443,742 B2 · Method and apparatus for determining a dialog state, dialog system, computer device, and storage medium > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,443,742 B2 - Document ID: US11443742B2 - Status: Granted - Kind: B2 - Application number: 16/921,300 - Filed: 2020-07-06 - Published: 2020-10-22 - Granted: 2022-09-13 - Earliest claimed: 2018-06-07 - Assignees: Tencent Technology (Shenzhen) Company Limited - Applicants: Tencent Technology (Shenzhen) Company Limited - Inventors: Xiaodong Lu - Web page: https://patentgrep.com/patent/US11443742B2 ## CPC - G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue - G06F 16/3329: Natural language query formulation - G06F 40/30: Semantic analysis - G06N 3/04: Architecture, e.g. interconnection topology - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/09: Supervised learning - G10L 15/1815: {Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning} - G06F 40/205: Parsing - G06F 40/35: Discourse or dialogue representation - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/08: Learning methods - G10L 15/1822: {Parsing for meaning understanding} - G10L 2015/228: {of application context} - Y02D 10/00: Energy efficient computing, e.g. low power processors, power management or thermal management - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE ## Abstract In a method for determining a dialog state, first dialog information is obtained. The first dialog information is dialog information inputted during a dialog process. Based on the first dialog information, target scenario information corresponding to the first dialog information is determined. The target scenario information is used to indicate a dialog scenario of the first dialog information. Based on the first dialog information and the target scenario information, a first dialog state corresponding to the first dialog information is obtained. The first dialog state is used to represent a response mode for responding to the first dialog information. ## Claims ### Claim 1 A method for determining a dialog state, the method comprising: obtaining, by circuitry of a computer device, first dialog information, the first dialog information being dialog information inputted during a dialog process; determining, by the circuitry of the computer device based on the first dialog information, target scenario information corresponding to the first dialog information, the target scenario information being used to indicate a dialog scenario of the first dialog information; determining, by the circuitry of the computer device based on a scenario vector of the target scenario information, at least one piece of candidate scenario information of a plurality of pieces of scenario information having an association relationship with the target scenario information; and obtaining, by the circuitry of the computer device based on the at least one piece of candidate scenario information and the target scenario information, a first dialog state according to the first dialog information, the first dialog state being used to represent a response mode for responding to the first dialog information, wherein the plurality of pieces scenario information is represented by a plurality of scenario vectors, each scenario vector representing an association relationship between the scenario information corresponding to the respective scenario vector and other scenario information of the plurality of pieces of scenario information, the plurality of scenario vectors is constructed according to at least one scenario network before the at least one piece of candidate scenario information is determined, the plurality of scenario vectors corresponding one-to-one to the plurality of pieces of scenario information, and the at least one scenario network is based on semantic tags of the plurality of pieces of scenario information, each semantic tag being a network node in the at least one scenario network and identifying a type of a dialog scenario indicated by the scenario information corresponding to the respective semantic tag. ### Claim 2 (depends on 1) The method according to claim 1, wherein before the determining the at least one piece of candidate scenario information, the method further comprises: obtaining, by the circuitry of the computer device, the semantic tags of the plurality of pieces of scenario information; building, by the circuitry of the computer device, the at least one scenario network based on the semantic tags of the plurality of pieces of scenario information; and constructing, by the circuitry of the computer device, the plurality of scenario vectors according to the at least one scenario network. ### Claim 3 (depends on 2) The method according to claim 2, wherein the constructing of the plurality of scenario vectors comprises: constructing, by the circuitry of the computer device, the plurality of scenario vectors according to a connection relationship between network nodes in the at least one scenario network. ### Claim 4 (depends on 3) The method according to claim 3, wherein the constructing of the plurality of scenario vectors comprises: walking, by the circuitry of the computer device along any path in the at least one scenario network to obtain a plurality of node sequences, each node sequence comprising at least two network nodes; determining, by the circuitry of the computer device, a similarity between each network node and another network node based on the plurality of node sequences; and constructing, by the circuitry of the computer device, the plurality of scenario vectors based on the similarity. ### Claim 5 (depends on 1) The method according to claim 1, wherein the determining of the target scenario information comprises: obtaining, by the circuitry of the computer device, a semantic understanding of the first dialog information based on the first dialog information; recognizing, by the circuitry of the computer device, a semantic tag of the first dialog information based on the semantic understanding; and querying, by the circuitry of the computer device, a scenario information set based on the semantic tag, to determine the target scenario information, the scenario information set including the plurality of pieces of scenario information. ### Claim 6 (depends on 5) The method according to claim 5, wherein the semantic tag of the first dialog information is a tag set comprising a plurality of semantic tags, and the querying of the scenario information set based on the semantic tag, to determine the target scenario information comprises: querying, by the circuitry of the computer device, the scenario information set separately based on the plurality of semantic tags, to obtain at least one of the plurality of scenario vectors corresponding to the plurality of semantic tags; and splicing, by the circuitry of the computer device when a plurality of the scenario vectors is found through querying, the plurality of found scenario vectors, to obtain the scenario vector of the target scenario information. ### Claim 7 (depends on 1) The method according to claim 1, wherein the determining the target scenario information comprises: querying a scenario information set based on a semantic tag of the first dialog information, and determining, by the circuitry of the computer device, first scenario information of the first dialog information; obtaining, by the circuitry of the computer device, second scenario information corresponding to second dialog information inputted before the first dialog information; and obtaining, by the circuitry of the computer device, the target scenario information based on the second scenario information and the first scenario information. ### Claim 8 (depends on 7) The method according to claim 7, wherein the obtaining of the target scenario information comprises: adjusting, by the circuitry of the computer device, the first scenario information based on the second scenario information, to obtain the target scenario information. ### Claim 9 (depends on 1) The method according to claim 1, wherein the obtaining of the first dialog state comprises: obtaining, by the circuitry of the computer device, a semantic understanding of second dialog information inputted before the first dialog information; and obtaining, by the circuitry of the computer device, the first dialog state based on the semantic understanding of the second dialog information, a semantic understanding of the first dialog information, and the target scenario information. ### Claim 10 (depends on 9) The method according to claim 9, wherein the obtaining of the first dialog state based on the semantic understanding of the second dialog information, the semantic understanding of the first dialog information, and the target scenario information comprises: inputting, by the circuitry of the computer device, the semantic understanding of the first dialog information and the target scenario information into a neural network, and determining, by the neural network, the first dialog state according to a second dialog state, the semantic understanding of the first dialog information, and the target scenario information, the second dialog state being determined according to the semantic understanding of the second dialog information. ### Claim 11 A computer device, comprising: circuitry configured to: obtain first dialog information, the first dialog information being dialog information inputted during a dialog process; determine, based on the first dialog information, target scenario information corresponding to the first dialog information, the target scenario information being used to indicate a dialog scenario of the first dialog information; determine, based on a scenario vector of the target scenario information, at least one piece of candidate scenario information of a plurality of pieces of scenario information having an association relationship with the target scenario information; and obtain, based on the at least one piece of candidate scenario information and the target scenario information, a first dialog state according to the first dialog information, the first dialog state being used to represent a response mode for responding to the first dialog information, wherein the plurality of pieces scenario information is represented by a plurality of scenario vectors, each scenario vector representing an association relationship between the scenario information corresponding to the respective scenario vector and other scenario information of the plurality of pieces of scenario information, the plurality of scenario vectors is constructed according to at least one scenario network before the at least one piece of candidate scenario information is determined, the plurality of scenario vectors corresponding one-to-one to the plurality of pieces of scenario information, and the at least one scenario network is based on semantic tags of the plurality of pieces of scenario information, each semantic tag being a network node in the at least one scenario network and identifying a type of a dialog scenario indicated by the scenario information corresponding to the respective semantic tag. ### Claim 12 (depends on 11) The computer device according to claim 11, wherein before the at least one piece of candidate scenario information is determined, the circuitry is configured to: obtain the semantic tags of the plurality of pieces of scenario information; build the at least one scenario network based on the semantic tags of the plurality of pieces of scenario information; and construct the plurality of scenario vectors according to the at least one scenario network. ### Claim 13 (depends on 12) The computer device according to claim 12, wherein in constructing the plurality of scenario vectors according to the at least one scenario network, the circuitry is configured to: construct the plurality of scenario vectors according to a connection relationship between network nodes in the at least one scenario network. ### Claim 14 (depends on 13) The computer device according to claim 13, wherein in constructing the plurality of scenario vectors the circuitry is further configured to: walk along any path in the at least one scenario network to obtain a plurality of node sequences, each node sequence comprising at least two network nodes; determine a similarity between each network node and another network node based on the plurality of node sequences; and construct the plurality of scenario vectors based on the similarity. ### Claim 15 (depends on 12) The computer device according to claim 12, wherein in determining the target scenario information corresponding to the first dialog information, the circuitry is configured to: query a scenario information set based on a semantic tag of the first dialog information, to determine first scenario information of the first dialog information; obtain second scenario information corresponding to second dialog information inputted before the first dialog information; and obtain the target scenario information based on the second scenario information and the first scenario information. ### Claim 16 (depends on 11) The computer device according to claim 11, wherein in determining the target scenario information the circuitry is configured to: obtain a semantic understanding of the first dialog information based on the first dialog information; recognize a semantic tag of the first dialog information based on the semantic understanding; and query a scenario information set based on the semantic tag, to determine the target scenario information, the scenario information set including the plurality of pieces of scenario information. ### Claim 17 (depends on 16) The computer device according to claim 16, wherein the semantic tag of the first dialog information is a tag set comprising a plurality of semantic tags, and in querying the scenario information set based on the semantic tag, the circuitry is further configured to: query the scenario information set separately based on the plurality of semantic tags, to obtain at least one of the plurality of scenario vectors corresponding to the plurality of semantic tags; and splice, when a plurality of the scenario vectors is found through querying, the plurality of found scenario vectors, to obtain the scenario vector of the target scenario information. ### Claim 18 A non-transitory computer-readable medium storing computer-readable instructions that, when executed by processing circuitry, cause the processing circuitry to perform a method comprising: obtaining first dialog information, the first dialog information being dialog information inputted during a dialog process; determining, based on the first dialog information, target scenario information corresponding to the first dialog information, the target scenario information being used for indicating a dialog scenario of the first dialog information; determining, based on a scenario vector of the target scenario information, at least one piece of candidate scenario information of a plurality of pieces of scenario information having an association relationship with the target scenario information; and obtaining, based on the at least one piece of candidate scenario information and the target scenario information, a first dialog state according to the first dialog information, the first dialog state being used for representing a response mode for responding to the first dialog information, wherein the plurality of pieces scenario information is represented by a plurality of scenario vectors, each scenario vector representing an association relationship between the scenario information corresponding to the respective scenario vector and other scenario information of the plurality of pieces of scenario information, the plurality of scenario vectors is constructed according to at least one scenario network before the at least one piece of candidate scenario information is determined, the plurality of scenario vectors corresponding one-to-one to the plurality of pieces of scenario information, and the at least one scenario network is based on semantic tags of the plurality of pieces of scenario information, each semantic tag being a network node in the at least one scenario network and identifying a type of a dialog scenario indicated by the scenario information corresponding to the respective semantic tag.