Knowledge representation learning method based on graph convolutional network and translation model

Abstract

A knowledge representation learning method based on a graph convolutional network and a translation model comprises the following steps: firstly, based on a knowledge base, learning entity and relation representation in the knowledge base by adopting the translation model; then, taking the knowledge base as guidance, and adopting remote supervision to obtain entities of the biomedical text and relation labels thereof; then using GCGCN to learn entity representation in the text; and finally, aligning entity representations in the knowledge base and the text, so that the entity representations based on knowledge base and remote supervised text learning coexist in the same vector space. Based on the translation model and the graph convolutional network, the knowledge base and large-scale remote supervision text information are effectively fused, multi-source information fusion is achieved, high-quality knowledge representation is obtained, and the performance of the biomedical relationship extraction model is improved. Structured knowledge in a knowledge base is learned based on a translation model, meanwhile, context knowledge in a large-scale remote supervision text is learned based on a graph convolutional network, and finally, multi-source knowledge is fused through entity alignment to obtain high-quality knowledge representation.

Description

technical field [0001] Based on Graph Convolutional Networks (GCN) and translation models, the present invention fuses triplets in the knowledge map and contexts in large-scale remote supervision texts to learn knowledge representation. First, based on the translation model, the knowledge representation is learned using knowledge base triples. Then, the graph convolutional network is used to learn the entities in the large-scale biomedical text obtained by distant supervision learning. Finally, entities in knowledge bases and biomedical texts are aligned to realize entity fusion based on knowledge bases and large-scale remote supervision text information. The invention is mainly used for the biomedical relationship extraction task in the field of natural language processing. Background technique [0002] With the rapid development of computer technology and biotechnology, the literature in the field of biomedicine is growing exponentially. Researchers are eager to reveal ...

AI Technical Summary

Problems solved by technology

This method fails to take all the sentences and entities in a document as a whole and learn the semantic representation of entities, entity relationships and sentences

Images