Reasearch On Joint Model Of Entity Recognition And Relation Extraction Based On Neural Network

With the advent of the information age,the information on the network is experiencing an explosive growth in the form of an index.Among them,text information occupies a very important part.How to accurately and efficiently acquire knowledge becomes an urgent problem.The primary goal of the joint model of entity recognition and relation extraction is to extract entity classes and semantic relationships between them from unstructured text.As the underlying technology of natural language processing technology,it has significant significance for the application of the upper layer.This paper focuses on the joint model of entity recognition and relationship extraction.The main work content and stage results are as follows:(1)The joint model baseline system based on neural network for entity recognition and relationship extraction is designed and implemented,and the possible problems of the current model are discussed.(2)A hybrid neural network structure based on parameter sharing and two-way long-term and short-term memory network-graph convolutional neural network is proposed.The model is used to better extract the relationships in sentences by introducing a syntactic convolutional neural network.On the public dataset,this federated model achieved better performance than before.(3)A joint model of fusion self-attention mechanism based on special tagging scheme is proposed.The information extraction task is transformed into sequence labeling task.The word dependency relationship in the sentence is learned through the self-attention sub-layer.The effect on the model's effect has also achieved good performance on the public data set.(4)A knowledge graph construction system based on entity recognition and relationship extraction joint model is built,which realizes the solution of extracting triples from unstructured text initially.