Research On Chinese Entity Relation Extraction Based On Rule Matching And Neural Network Learning

Entity Relation Extraction(ERE)is an important research problem in the field of information extraction.It can recognize the semantic relation between entity pairs.This technology plays an important role in knowledge graph construction,question answering system,and semantic search and other fields.Most of the methods in the research of ERE are based on English corpus.Because the syntactic structure features of Chinese and English sentences are very different,the existing English corpus-based methods are difficult to apply directly to Chinese corpus,and the Chinese datasets of this research is relatively lacking.Therefore,the research on ERE for Chinese text has important theoretical significance and broad application prospects.The research work in this dissertation is as follows:(1)Chinese entity relation extraction based on syntactic features.This dissertation proposes a novel Chinese ERE method based on syntactic features.Firstly,the candidate relation triples are extracted based on verbs and nouns as relation keywords to avoid pre-defining relation types.Secondly,the triples are filtered using the positional constraints between relation keywords and entity pairs.Thirdly,expand the identified relation triples by summarizing four major Chinese syntactic features and the accuracy of ERE is improved.Finally,use the method of relation transfer to mine and infer implicit relation triples.The experimental results on the encyclopedia dataset and the news dataset show that the method has good extraction performance.(2)Chinese personal relation extraction based on trigger word rules and Att-BLSTM.Due to the lack of training data in Chinese field,the Chinese ERE method based on neural network model is less and the accuracy is lower.For this reason,the training data is expanded by the extraction algorithm based on trigger word rules,thereby improving the accuracy of the neural network model.In further,the effect of personal relation extraction is improved.The extraction algorithm based on trigger word rules realizes the automatic labeling of training data,and forms a small-scale annotated Chinese personal relation extraction dataset.The experimental results show that the Chinese personal relation extraction method combined with trigger word rules and Att-BLSTM is superior to the baseline method based on Att-BLSTM.