Entity linking method based on integration of knowledge graph subgraph information and entity information

Abstract

The invention relates to an entity linking method based on integration of knowledge graph subgraph information and entity information, and the method comprises the following steps: obtaining a question text, feeding the question text into a trained mention recognition model, and recognizing a prediction mention word; recalling the predicted mentions to obtain target mentions with higher accuracy; obtaining all candidate entities corresponding to the mentions from a knowledge base according to the target mentions; for each candidate entity, retrieving knowledge graph sub-graph information of the candidate entity; integrating knowledge graph sub-graph information into candidate entities to obtain candidate entity texts, and introducing entity boundary tags into the question text and the candidate entity texts to construct an entity disambiguation data set; and feeding a trained entity disambiguation model, pooling to obtain a knowledge base entity corresponding to the mention, and completing entity linking. According to the method, the recall rate of mention word recognition can be effectively improved, the sorting accuracy of candidate entities is improved, error transmission is reduced, and the problems of reasoning and no entity in a training set are effectively solved.

Description

technical field [0001] The present invention relates to the field of natural language processing of artificial intelligence, in particular to an entity linking method based on incorporating knowledge map subgraph information and entity information. Background technique [0002] Knowledge base question answering (KBQA) is a given natural language question, through the semantic understanding and analysis of the question, and then use the knowledge base to query and reason to get the answer. There are two types of methods: methods based on semantic analysis (SP) and methods based on information extraction (IE). Among them, the information extraction method has a relatively fixed model framework and better generalization performance. This type of method generally adopts a pipeline architecture, and first performs (1) entity linking, that is, linking the mention word in the question to the entity of the knowledge base. Then (2) relation identification is carried out, that is, t...

AI Technical Summary

Problems solved by technology

[0004] Based on this, in order to solve the problem of small amount of information in the upper text and blurred boundaries of Chinese named entities, the present invention provides an entity link method based on the integration of knowledge map subgraph information and entity information, by improving the performance of KBQA entity link tasks, and then Further improve the performance of the whole knowledge base question answering task to address the problems mentioned in the background above

Images