# US 12,216,697 B2 · Event detection > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,216,697 B2 - Document ID: US12216697B2 - Status: Granted - Kind: B2 - Application number: 18/395,120 - Filed: 2023-12-22 - Published: 2024-05-02 - Granted: 2025-02-04 - Earliest claimed: 2021-09-14 - Assignees: Alipay (Hangzhou) Information Technology Co., Ltd. - Applicants: Alipay (Hangzhou) Information Technology Co., Ltd. - Inventors: Weipeng Huang - Web page: https://patentgrep.com/patent/US12216697B2 ## CPC - G06F 16/358: Browsing; Visualisation therefor - G06F 40/30: Semantic analysis - G06F 16/367: {Ontology} - G06F 16/951: Indexing; Web crawling techniques - G06F 40/40: Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30) - G06F 40/211: Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract This specification discloses an event detection method and system. The method includes: obtaining a to-be-processed text; extracting one or more groups of instance data from the to-be-processed text by using an extraction model; determining one or more extracted triplets based on the one or more groups of instance data, to obtain an extracted graph; obtaining graph ontology definition data of one or more candidate events, and obtaining an ontology definition graph corresponding to each candidate event based on the graph ontology definition data; determining a similarity between the extracted graph and each ontology definition graph of the one or more candidate events; and determining an event corresponding to the to-be-processed text from the one or more candidate events based on each similarity. ## Claims ### Claim 1 An event detection method, the method comprising: obtaining a to-be-processed text; extracting one or more groups of instance data from the to-be-processed text by using an extraction model, each group of instance data including a first entity instance, a first entity type corresponding to the first entity instance, a second entity instance, a second entity type corresponding to the second entity instance, and a relationship description between the first entity type and the second entity type; determining one or more extracted triplets based on the one or more groups of instance data, to obtain an extracted graph, an extracted triplet including the first entity type, the second entity type, and the relationship description between the first entity type and the second entity type in the instance data; obtaining graph ontology definition data of one or more candidate events, and obtaining an ontology definition graph corresponding to each candidate event based on the graph ontology definition data, the graph ontology definition data of the one or more candidate events including an entity type used to indicate an entity and a relationship description used to indicate a relationship between entity types; determining a similarity between the extracted graph and each ontology definition graph of the one or more candidate events; and determining an event corresponding to the to-be-processed text from the one or more candidate events based on each similarity. ### Claim 2 (depends on 1) The method according to claim 1, further comprising: determining the graph ontology definition data of events, wherein entity type and relationship description of the graph ontology definition data of the events are determined from an ontology definition data set. ### Claim 3 (depends on 2) The method according to claim 2, wherein the extraction model is obtained through training based on a training sample, and a label of the training sample is determined based on entity type and relationship description in the ontology definition data set. ### Claim 4 (depends on 1) The method according to claim 1, wherein the relationship description includes one or more of: a verb-object relationship, a subject-verb relationship, an attribute-object relationship, or a modifying relationship. ### Claim 5 (depends on 1) The method according to claim 1, wherein the extracting the one or more groups of instance data from the to-be-processed text by using the extraction model includes: processing the to-be-processed text by using the extraction model, to obtain an annotation sequence and a relationship matrix of the to-be-processed text; determining an entity instance and an entity type of an entity instance in the to-be-processed text based on the annotation sequence; and determining a relationship description between two entity instances in the to-be-processed text based on the relationship matrix, and using the relationship description as a relationship description between entity types corresponding to the two entity instances. ### Claim 6 (depends on 5) The method according to claim 5, wherein the annotation sequence is used to mark a word or a phrase in the to-be-processed text that is an entity instance and an entity type of the word or the phrase, and the relationship matrix is used to mark a relationship description between two of words or phrases in the to-be-processed text. ### Claim 7 (depends on 1) The method according to claim 1, wherein the extraction model includes one or more of: BERT, Transformer, StanfordNLP, or LTP. ### Claim 8 (depends on 1) The method according to claim 1, wherein the determining the similarity between the extracted graph and each ontology definition graph of the one or more candidate events includes: for an ontology definition graph of a candidate event, processing the extracted graph and the ontology definition graph of the candidate event by using a graph matching model, to obtain a similarity between the extracted graph and the ontology definition graph of the candidate event. ### Claim 9 (depends on 1) The method according to claim 1, further comprising: determining an event element of an event corresponding to the to-be-processed text based on the one or more instance triplets. ### Claim 10 A computer system comprising at least one processor and at least one memory, the at least one memory, individually or collectively, having computer executable instructions stored thereon, which when executed by the at least one processor, enabling the at least one processor to, individually or collectively, implement acts including: obtaining a to-be-processed text; extracting one or more groups of instance data from the to-be-processed text by using an extraction model, each group of instance data including a first entity instance, a first entity type corresponding to the first entity instance, a second entity instance, a second entity type corresponding to the second entity instance, and a relationship description between the first entity type and the second entity type; determining one or more extracted triplets based on the one or more groups of instance data, to obtain an extracted graph, an extracted triplet including the first entity type, the second entity type, and the relationship description between the first entity type and the second entity type in the instance data; obtaining graph ontology definition data of one or more candidate events, and obtaining an ontology definition graph corresponding to each candidate event based on the graph ontology definition data, the graph ontology definition data of the one or more candidate events including an entity type used to indicate an entity and a relationship description used to indicate a relationship between entity types; determining a similarity between the extracted graph and each ontology definition graph of the one or more candidate events; and determining an event corresponding to the to-be-processed text from the one or more candidate events based on each similarity. ### Claim 11 (depends on 10) The computer system according to claim 10, wherein the acts further include: determining the graph ontology definition data of events, wherein entity type and relationship description of the graph ontology definition data of the events are determined from an ontology definition data set. ### Claim 12 (depends on 11) The computer system according to claim 11, wherein the extraction model is obtained through training based on a training sample, and a label of the training sample is determined based on entity type and relationship description in the ontology definition data set. ### Claim 13 (depends on 10) The computer system according to claim 10, wherein the relationship description includes one or more of: a verb-object relationship, a subject-verb relationship, an attribute-object relationship, or a modifying relationship. ### Claim 14 (depends on 10) The computer system according to claim 10, wherein the extracting the one or more groups of instance data from the to-be-processed text by using the extraction model includes: processing the to-be-processed text by using the extraction model, to obtain an annotation sequence and a relationship matrix of the to-be-processed text; determining an entity instance and an entity type of an entity instance in the to-be-processed text based on the annotation sequence; and determining a relationship description between two entity instances in the to-be-processed text based on the relationship matrix, and using the relationship description as a relationship description between entity types corresponding to the two entity instances. ### Claim 15 (depends on 14) The computer system according to claim 14, wherein the annotation sequence is used to mark a word or a phrase in the to-be-processed text that is an entity instance and an entity type of the word or the phrase, and the relationship matrix is used to mark a relationship description between two of words or phrases in the to-be-processed text. ### Claim 16 (depends on 10) The computer system according to claim 10, wherein the extraction model includes one or more of: BERT, Transformer, StanfordNLP, or LTP. ### Claim 17 (depends on 10) The computer system according to claim 10, wherein the determining the similarity between the extracted graph and each ontology definition graph of the one or more candidate events includes: for an ontology definition graph of a candidate event, processing the extracted graph and the ontology definition graph of the candidate event by using a graph matching model, to obtain a similarity between the extracted graph and the ontology definition graph of the candidate event. ### Claim 18 (depends on 10) The computer system according to claim 10, wherein the acts further include: determining an event element of an event corresponding to the to-be-processed text based on the one or more instance triplets. ### Claim 19 A non-transitory storage medium having computer executable instructions stored thereon, the computer executable instructions when executed by one or more processors, enabling the one or more processors to, individually or collectively, implement acts comprising: obtaining a to-be-processed text; extracting one or more groups of instance data from the to-be-processed text by using an extraction model, each group of instance data including a first entity instance, a first entity type corresponding to the first entity instance, a second entity instance, a second entity type corresponding to the second entity instance, and a relationship description between the first entity type and the second entity type; determining one or more extracted triplets based on the one or more groups of instance data, to obtain an extracted graph, an extracted triplet including the first entity type, the second entity type, and the relationship description between the first entity type and the second entity type in the instance data; obtaining graph ontology definition data of one or more candidate events, and obtaining an ontology definition graph corresponding to each candidate event based on the graph ontology definition data, the graph ontology definition data of the one or more candidate events including an entity type used to indicate an entity and a relationship description used to indicate a relationship between entity types; determining a similarity between the extracted graph and each ontology definition graph of the one or more candidate events; and determining an event corresponding to the to-be-processed text from the one or more candidate events based on each similarity. ### Claim 20 (depends on 19) The non-transitory storage medium according to claim 19, wherein the extraction model is obtained through training based on a training sample, and a label of the training sample is determined based on entity type and relationship description in the ontology definition data set.