A method and device for semantic understanding and precise matching of short texts based on knowledge graph

Abstract

The invention discloses a short text semantic understanding and precise matching method based on a knowledge map, which extracts the first entity from the first text input by the user, finds all candidate entities corresponding to the first entity from the built-in knowledge map, and extracts The first candidate entity with the highest confidence, and then find the first knowledge set corresponding to the first entity according to the built-in knowledge map, and find the second knowledge set corresponding to the first candidate entity in the knowledge base, and combine the first knowledge set and the second knowledge set The knowledge set is input into the LSTM intent recognition classification model of semantic understanding and the LSTM+CRF knowledge sequence labeling model to generate the first knowledge feature and the second knowledge feature, and combine the two to generate multiple sets of combined features, and input them into the logistic regression The text matching model performs matching judgment, generates texts corresponding to all matching combined features, and takes the text with the highest confidence as the best matching text, which can achieve better semantic understanding and make the matching process more accurate.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to a short text semantic understanding and precise matching method based on a knowledge graph. Background technique [0002] Short text semantic understanding and precise matching is a natural language processing technology widely used in human-computer dialogue systems. After semantic understanding of short texts (including intent recognition and attribute slot extraction), it can be compared with specific texts in the knowledge base. Matching, find the most similar preset text set, so as to get the corresponding reply, knowledge and other data. [0003] But existing technology has following shortcoming: [0004] (1) Short texts are difficult to understand semantically due to the short text and lack of context; [0005] (2) When short texts are matched with specific texts in the knowledge base, there is a lack of knowledge support and it is difficult to ac...

AI Technical Summary

Problems solved by technology

[0004] (1) Short texts are difficult to understand semantically due to the short text and lack of context;

[0005] (2) When short texts are matched with specific texts in the knowledge base, there is a lack of knowledge support and it is difficult to achieve accurate matching

Images