Research On The Construction And Application Of Knowledge Graph In The Field Of Culture And Entertainment

With the continuous expansion of the online market of culture and entertainment,people’s pursuit of spiritual culture and quality of life is rising,and the society’s demand for applications in the field of culture and entertainment is also getting higher and higher.Faced with the increasing number of entertainment information,knowledge graph,with its characteristics of structure and network,can better represent the correlation between entertainment information,and help users to understand the connection between entertainment information more intuitively.At present,there are more and more applications of knowledge graph for specific fields,but there are few applications of knowledge graph related to the field of culture and entertainment in the market,and there is also a lack of open source knowledge graph in the field of culture and entertainment.At the same time,as a new direction of the future development of search engines,the application of knowledge graph in intelligent question-answering can infer more accurate answers according to the user’s intention,which has gradually become a new trend of human-machine interaction.In view of the above background,this thesis uses the knowledge in the field of culture and entertainment to construct a knowledge graph in the field of culture and entertainment,designs a question-answer model oriented to the knowledge graph in the field of culture and entertainment,and builds a Web-based question-answer retrieval platform for the knowledge graph in the field of culture and entertainment.Specific contents include:(1)Construction of knowledge graph in the field of culture and entertainment.In this thesis,knowledge graph in the field of culture and entertainment is constructed by knowledge extraction,entity alignment and knowledge storage,which includes 132284 entities and 388552 relationships.In order to solve the problem of error propagation in entity and relationship extraction,a joint entity and relationship extraction model based on BERT language model is designed in this thesis.At the same time,aiming at the possibility of multiple triples in a single text,this thesis applies the idea of pointer combined annotation to the joint extraction model of entity and relationship.In order to solve the problem that a single similarity calculation method cannot make full use of text features,this thesis designs a method of entity alignment combining Jaccard coefficient and edit distance,and fuses entities obtained from different data sources.For the two kinds of culture and entertainment data extracted,this thesis uses Neo4 j graph database and Mongo DB document database for knowledge storage respectively.(2)Question-and-answer model for knowledge graph in the field of culture and entertainment.In this thesis,the question-and-answer model based on knowledge graph in the field of culture and entertainment is realized through entity identification,candidate answer generation and entertainment interrogative-relational semantic matching model.Aiming at the problem that manual annotation of a large number of question-and-answer model data requires a large amount of manpower and time costs,this thesis integrates the open domain question-and-answer corpus with the question-and-answer corpus in the culture and entertainment field,and constructs the experimental data set of the question-and-answer model in this thesis.In view of the importance of entity recognition task in question and answer model,this thesis uses the BERT-BiLSTM-CRF model to extract entities from questions.Through comparative experiments,it is proved that this model can effectively improve the effect of entity recognition.Aiming at the disadvantage of using only the vector representation of [CLS] position in the downstream BERT task,this thesis combined onedimensional convolution and maximum pooling operation to design a playful interrogationrelation semantic matching model based on BERT language model,which realized the fusion of all the coding information of BERT coding sequence,and improved the ability of the model to identify similar relationship.(3)A Web-based Q&A retrieval platform for knowledge graph in the field of culture and entertainment.Based on the constructed knowledge graph in the field of culture and entertainment and the designed question-answer model of knowledge graph in the field of culture and entertainment,this thesis builds a Web-based Q&A retrieval platform for knowledge graph in the field of culture and entertainment through technologies such as Flask development framework,D3.js graphic visualization framework and Bootstrap front-end visualization framework.The platform is accessed in the form of web pages,and provides users with functions such as visualization,retrieval,expansion and question-and-answer of knowledge graph in the field of culture and entertainment,which meet people’s daily needs for applications in the field of culture and entertainment.