Question and answering method for embedding multiple knowledge maps in combination with hyperbolic segmented knowledge of text

Abstract

A question answering method for embedding multiple knowledge maps in combination with hyperbolic segmented knowledge of a text is characterized in by comprising the following steps: using a hyperbolic segmented embedding model to realize initialization of entities and relationships in the knowledge maps, wherein the entities comprise a head entity and a tail entity; aiming at a problem proposed by a natural language, performing problem embedding by utilizing a RoBERTa model, constructing a heterogeneous graph, and fusing a knowledge graph and a text related to the problem; utilizing the semantic information in the text to enrich the expression of entity embedding in the knowledge graph, and using the text as a hyperedge to supplement the relationship in the incomplete knowledge graph; and constructing a new triple by combining entity embedding, question embedding and candidate answers of the text, and scoring by adopting a scoring function to realize knowledge graph question and answer. The quality of knowledge graph questions and answers can be greatly improved, and more accurate answers can be obtained.

Description

Technical field [0001] The present invention belongs to the technical field of artificial intelligence, to mapping knowledge quiz, particularly relates to a method of binding text Q hyperbolic segment embedding of multiple knowledge mapping of knowledge. Background technique [0002] In recent years, large-scale development of the knowledge map provides a wealth of resources for open-domain question answering. Based on mapping knowledge to answer natural language questions has become a popular trend. [0003] Most of the early mapping knowledge quiz uses a traditional approach based on semantic analysis. Such methods use a dictionary, rules and machine learning, natural language the question is mapped into a semantic representation or logical expression or query graph. Commonly used method to resolve semantic-based classification model to predict the relationship, but because of the knowledge map contains hundreds of thousands of relationships, the relationship between the traini...

AI Technical Summary

Problems solved by technology

However, how to truly understand the problem and bridge the gap between natural language and the structured semantics of knowledge graphs is still very challenging.

[0005] In addition, in knowledge graph question answering based on neural networks, it is usually necessary to learn the representation of entities and relations in triples through knowledge embedding models. Make an appropriate trade-off between model complexity (number of parameters) and model expressivity (semantic information capture performance), and at the same time, it is not easy to maintain the multi-level structure of the knowledge graph

Images