Research And Implementation Of Question Answering System Based On Knowledge Graph

With the explosive growth of information on the Internet,it is difficult for traditional search engines to meet needs of users for precise information retrieval.The question answering system classifies the questions semantically,and then return the corresponding answers more accurately and intuitively,which improves user experience.However,most question answering systems are constructed based on rules,which are less flexible and require more human intervention.In order to improve the above defects and shortcomings of the question answering system,this paper uses the knowledge graph as the knowledge base,and focuses on the information extraction part,and proposes the BERT-BiLSTM-MA-CRF named entity recognition model.On the one hand,the model uses the BERT language model as the word embedding layer,which simplifies the feature extraction process and solves the problem of polysemy in information extraction.On the other hand,the model integrates the multi-head attention mechanism to capture the sequence structure.The experimental results show that the model has achieved good results in the named entity recognition task,and it still needs to be optimized on the nested entity recognition task.To address the problems in the nested entity recognition task,this paper proposes the KE-BTBMC model,which embeds the knowledge representation into the model and stacks the NER layers for optimization,which improves the limitations of the non-nested named entity recognition model in the nested entity scene.Experiments show that compared with mainstream named entity recognition models such as BiLSTM-CRF,the model achieves significant improvement in F1 value on nested entity datasets,and it still has good applicability on non-nested entity datasets.Finally,based on the above model,in the medical field which exists many nested entities,this paper constructs a question answering system based on knowledge graph.The system can analyze the questions input by users,query the knowledge base and return corresponding answers,helping people faster more accurate access to medical knowledge.Therefore,the system has certain practicality.