Research On Question Answering Based On Knowledge Base

With the development of artificial intelligence technology,the automatic question answering technology has received extensive attention.Unlike traditional search engines,the main goal of the question answering system is to return correct and concise answers to the user,rather than a collection of a large number of candidate answers.The knowledge base is the basis of text resources.By using the knowledge base,the semantic knowledge contained in text resources can be analyzed in detail.In the research of knowledge quiz technology for knowledge base in this topic,it mainly includes the following three aspects: knowledge base question answering technology based on single-relation information retrieval,knowledge base question answering technology based on multi-relation information retrieval and knowledge base based on semantic parsing.For the knowledge base question answering technology based on singlerelation information retrieval,this topic mainly explores the relationship recognition technology.The previous work mainly considered how to establish the matching method between the question and the question,and ignored the importance of the syntax in the question.Therefore,this paper attempts to use the dependency tree to propose a relationship recognition model based on the bidirectional tree structure.The experimental results show that the method proposed in this paper has a significant improvement over the previous work.Aiming at the knowledge base question answering technology based on multirelation information retrieval,this topic adds the constraint recognition technology based on the single-relation information retrieval knowledge base question answering technology,which constitutes an end-to-end knowledge base question answering system.The experimental results show that the knowledge base question answering system constructed by this topic has a significant effect.In the end,we have carried out a detailed analysis of the errors.Aiming at the knowledge base question answering technology based on semantic parsing,this thesis first introduces the sequence-to-action model,defines the semantic graph and action sequence,and proposes a semantic analysis model based on the copy mechanism and a semantic-analytical model of executionoriented based on the sequence-to-action model.The experimental results show that the two methods have a significant improvement compared to the baseline method based on the sequence-to-sequence semantic analysis model.In the end,we compare the information retrieval-based model and the semantic parsing-based model mentioned in this topic in detail,compare the differences between both,and explain the influence of different factors on the two models.The advantages and disadvantages of the two models are generally explained.