Research On Key Techniques Of Named Entity Recognition And Relationship Extraction

Knowledge Graph is a structured semantic knowledge base,which describes concepts and relations in the physical world with the form of symbols.It can realize concept retrieval through reasoning,and graphically display structured knowledge after sorting.In recent years,with the development of technological innovation,Knowledge Graph has been applied in various fields as an intelligent and efficient knowledge organization.Named Entity Recognition,the key link of constructing knowledge graph,means to identify special entity from the text dataset automatically.Currently,the named entity recognition technology mostly uses a hybrid model of deep learning and statistical methods to identify named entities.The current methods have reached more than 90%in effect,but there are still many challenges in Chinese Named Entity Recognition.For the issues,this paper proposes a named entity recognition method based on the BiLSTM-CCRFs model.This method combines the Bi-LSTM and the CCRFs,which not only reduces the manual participation,but also improves the effect of complex structured entities recognition.Entity Relation Extraction is a deeper research based on Named Entity Recognition,which means to extract relationships between entities from text datasets automatically.In recent years,entity relation extraction has gradually transitioned from pipeline methods based on named entity recognition to joint extraction methods based on deep learning.At present,there are two joint extraction methods: parameter sharing and annotation strategy.This paper proposes a joint entity relation extraction method based on annotation strategy.This method achieves joint extraction through the Chinese sequence annotation strategy and adopts the attention-based Bi-LSTM to complete the end-to-end triple relation extraction.To verify the feasibility of the proposed methods,this paper applies theory to practice and builds a knowledge graph construction system.This system includes named entity recognition module,entity relation extraction module,knowledge retrieval module and models management module.By default,the named entity recognition module invokes the Bi-LSTM-CCRFs named entity recognition model,and the entity relation extraction module invokes the joint relation extraction model.The knowledge retrieval module provides a knowledge retrieval service based on the system database to users and displays the results in a graphical format.