Research And Implementation Of Domain Question Answering System Based On Knowledge Graph

As one of the most important tasks in the field of natural language processing,the question and answer system provides people with a more intelligent,efficient and convenient information retrieval method,and is also one of the important directions of artificial intelligence research.Because of the large differences in domain knowledge and the high degree of professional knowledge in various industries,the question and answer system in the construction field has more practical value and significance.At the same time,with the development of knowledge map technology,knowledge map has the advantages of providing high credibility and complete knowledge of related information.Therefore,the question and answer system based on knowledge map has become the focus of researchers.This paper designs and implements a domain question answering system based on knowledge graph based on question entity recognition,attribute classification and knowledge base retrieval algorithm.The main works of this paper are as follows:1.It is difficult to accurately delimit the physical boundary of the entity referential form included in the question.This paper proposes a question-based entity recognition method based on feature fusion.The question is firstly expressed by the BERT model,and then the question features are extracted and merged through the BiLSTM and CNN networks.Finally,the conditional random field is input to further learn the label constraints,thereby improving the accuracy of the entity recognition.2.For the attribute extraction stage,it is difficult to understand the true meaning of the question only through the semantic matching of<question,attribute>.This paper proposes to introduce the answer information to assist in the answer screening.It is composed of<question,attribute>and<question,answer>,and it is combined semantic representation by BERT model.At the same time,in order to assign the weight of attribute and answer to the degree of question relevance,this paper proposes to adopt self-attention mechanism.Automatically assign weights.3.This paper combines multiple sources by constructing CIDOC CRM ontology model and vocabulary.According to this,a knowledge base of cultural relics field is constructed,and a domain knowledge base question answering system is designed based on the knowledge base design.