A multi-intent recognition system, method, device and medium based on knowledge graph

Abstract

The present invention discloses a multi-intent recognition system, method, device and medium based on a knowledge graph, including: a knowledge graph, used to provide data basis for graph calculation and reasoning; an entity recognition module, used to, based on the knowledge graph, from The text entity and text attribute are identified in the user input text; the graph calculation module is used to perform graph calculation based on the knowledge map, the text entity and the text attribute output by the entity recognition module, and identify user problems; the interaction module is used for Call the graph calculation module to obtain the user questions, and recall and generate answers from the knowledge base. The present invention automatically performs multi-intent recognition based on the knowledge map, not limited to punctuation, sentence patterns and syntax analysis. When the user's intention is unclear, the intention convergence is realized through automatic rhetorical questioning, which effectively improves the question recognition rate and accuracy rate, and greatly improves the performance of the question-and-answer robot. Flexibility, natural conversation.

Description

technical field [0001] The present invention relates to the technical field of computer question answering, in particular to a knowledge map-based multi-intent recognition system, method, device and medium. Background technique [0002] At present, in the field of question answering, the traditional multi-intent recognition uses simple sentence segmentation based on punctuation, or combined with syntactic analysis, such as "what is justifiable defense? How many years is the sentence for excessive defense?" Divide it into two questions: "What is justifiable self-defense" and "How long is the sentence for excessive defense?" According to the divided questions, they are recalled in the knowledge base through similar matching; another example is "Is murder illegal? How long is the sentence?" " can be segmented into "Is murder illegal?" and "How long is the sentence?" At this time, through syntactic analysis of the segmented questions, it can be seen that the latter sentence lack...

AI Technical Summary

Problems solved by technology

It can be seen that the existing technology requires the questioner to be able to use correct grammatical expressions and even correct punctuation for multi-intent recognition. More missing content, more grammatical errors, and the intention of the question is unclear, and it is difficult to correctly identify the intention through punctuation and sentence syntax analysis

Images