Research On Entity Relation Extraction Method Based On Deep Learning

Entity relation extraction is one of the most important research directions in the field of natural language processing,aiming to identify and judge the specific relationship between entities from unstructured text.With the rapid development of the Internet,a large amount of unstructured text information is produced,which contains rich and valuable information,and people have more urgent needs for data information extraction.Therefore,the technology of entity relation extraction has been widely concerned in industry and academia.The construction of efficient and accurate entity relationship extraction model is not only the basis of text mining,knowledge graph construction,information retrieval and other applications,but also can play a huge role in vertical fields such as e-commerce,finance,medical treatment and so on.Traditional entity relation extraction methods usually adopt a pipeline-based approach,which divides entity recognition and relation extraction into two independent sub-tasks.Such a model has the problems of information redundancy,error propagation and lack of dependency between the two sub-tasks.In recent years,the joint extraction model of entity relation based on deep learning has fully demonstrated its advantages and gradually became the mainstream method at present.This paper studies the method of extracting the joint entity relationship based on deep learning,and the main innovations are as follows:A joint entity relation extraction method based on double pointer network is proposed.In order to make full use of the entity information,this paper designs a dual pointer network module,which can copy the complete entity by predicting the beginning and the end position of the entity in the sentence.In order to eliminate the effect of triplet extraction order on the result of entity relation extraction,the reinforcement learning training strategy is adopted to improve the performance of the model.Full experiments have been carried out on the public datasets NYT and Web NLG,and F1 is worthy of effective improvement.A joint entity relation extraction method based on graph convolutional network is proposed.Aiming at the long-standing entity nesting problem,a boundary detection module is designed to improve the accuracy of entity recognition.In order to solve the overlapping problem of relations,a complete entity-relation graph was constructed by the proposed graph convolutional network,which enhanced the interactivity of the span of entities and relations,and further excavated the hidden features.In the public data set ACE05,the effect of entity recognition has been significantly improved.On the NYT and Web NLG common datasets,the model in this paper can effectively solve the problem of relationship overlap,and the extraction effect of entity relationship is significantly improved compared with the baseline model.