Research And Implementation Of Automatic Question Answering System Based On Disease Knowledge Graph Of Imaging Department

Medical imaging is the main means of early diagnosis of cancer.With the increasing incidence of cancer in China,the demand for medical imaging and disease prevention knowledge is becoming more and more urgent.Search engines lack the understanding of users’ intentions,and the existing online consultation services rely on manual work,which can not ensure immediate responses.Automatic question answering can effectively solve the above problems,by using the knowledge map to realize the organization and management of knowledge,and providing data support for accurate question answering.However,there is still a lack of complete knowledge map in the vertical field of imaging diseases,and in the process of questioning and answering,and the semantic analysis based on deep learning is also limited by the accuracy of the algorithm,which is a great challenge.In this thesis,the semantic analysis question and answer is combined with the recommendation question and answer,and personalized question recommendation makes up for the deficiency of semantic analysis and improves the question and answer service.Aiming at the key technology of named entity recognition in semantic parsing question answering,aiming at and the problem that the character based BiLSTM-CRF model will ignore the word information in the question sentence,and it is difficult to recognize rare words in the medical field,this thesis proposes CRSD-BiLSTM-CRF model,which integrates the internal structural features of Chinese characters and dictionary information into the model embedding layer.The experimental results show that the F1 value of CRSD-BiLSTM-CRF model is improved by 3.89%,and the accuracy is also higher.Aiming at the problem of data sparseness in recommendation question and answer,a hybrid collaborative filtering recommendation algorithm HCFDS for sparse data is proposed.Aiming at the significant deviation in similarity calculation in the data sparse environment,the score difference threshold and the same score penalty factor are introduced to solve the problem.Aiming at the data missing problem in the score matrix,the project category preference score is integrated into the matrix filling optimization.Experiments show that HCFDS algorithm can effectively solve the problem of data sparsity.This thesis constructs a complete image disease knowledge map.On this basis,using the above algorithm research results,an automatic question answering system based on disease knowledge graph of imaging department is designed and implemented.This thesis first introduces the research background and research status,then gives an overview of related technologies,then presents an analysis of the needs of the system,and then introduces the research on medical named entity recognition algorithm based on character multi features and hybrid collaborative filtering recommendation algorithm for data sparsity,which solves the key problems of this paper.Then,the outline design of the system is carried out,and the division of each subsystem and its functional modules is completed.Then it introduces the detailed design and realization of the system.Finally,the function test and performance test of the system are implemented.