The Construction And Retrieval Of Knowledge Graph For The Biographical History Books

A large number of historical books have been circulating in China.Since the New Culture Movement promoted vernacular writing,it has gradually made it more difficult for people to read classical Chinese books.In order to avoid the situation that no one can understand,the experts of the national organization rescued the translation of "Twenty-Four History",but there are still problems such as scattered narrative,huge space,many characters,and complicated relationships,etc.,in the chronicle history books.Knowledge graph has strong semantic representation ability and good expansion ability,which makes it have good performance in knowledge organization,knowledge retrieval and knowledge display,etc.It provides low-level data and technical support for applications such as intelligent question answering,intelligent search and recommendation.Fine-grained knowledge extraction of historical information,mainly based on time,place,and people,supplemented by other elements to build a relational semantic web,and stored in the form of knowledge graph,will help integrate historical event records and improve retrieval efficiency.The specific research work of this article is divided into two parts.The first part is to complete the construction of historical information knowledge graph.First of all,by analyzing the original data of the Biographical History Book,the ontology model of historical information is constructed,and the method of knowledge acquisition in the form of quadruple is designed.Secondly,the BiLSTM + CRF model was used to complete the named entity recognition task,and a good recognition effect was achieved.Thirdly,in order to save more fine-grained entity relations,a graph decoding dependency syntax analysis model combining semantic features and global features is designed to complete the task of expressing entity relations in terms of related words recorded in books.Finally,according to the history book of history,a storage scheme is designed,and the construction of knowledge graph is completed using neo4 j graph database.The second part is to propose a semantic-based retrieval model for the constructed historical information knowledge graph.First,the semantic triples are extracted through word segmentation and syntactic analysis,and then the semantic triples are constructed into a Cypher query language based on the sample set of question sentences.If no results are returned,perform semantic analysis and search again.Through the contrast experiment,the accuracy of named entity recognition model is 82.5%,recall rate is 83.3%,F value is 82.9%,dependency relationship accuracy(UAS)is 86.5%,dependency type accuracy(LAS)is 83.4%,and retrieval sentence classification accuracy is 84.5%.The feasibility of named entity recognition model,entity relationship extraction model and semantic retrieval model is verified.