A Biomedical Question Answering System Based On Biomedical Ontology And Multiple Mapping Templates

With the accelerated pace of life and the advent of an aging society,health has become a problem that people are increasingly concerned about,and people’s demand for medical knowledge has gradually increased.Moreover,physicians often have a lot of clinical problems and need some extra help.Therefore,it is of great significance to study the medical question answering system to answer the clinical questions of medical personnel and enable people to acquire medical knowledge more conveniently.However,the research on medical question answering system based on semantic technology is still in its infancy.Some medical question answering systems based on semantic technology are still insufficient in semantic analysis and processing for user questions,and the accuracy of information retrieval needs to be improved.To solve these problems,this paper proposes a biomedical question answering system based on biomedical ontology and multiple mapping templates to answer the clinical questions posed by physicians in natural language form.The system uses the semantic network in SNOMED-CT biomedical ontology as knowledge base,and the multiple mapping question template based on description logic is adopted to map the user questions into the different semantic relationships in SNOMED-CT,so as to mostly satisfy users’ needs to access to the information related to the problem.Moreover,a textual entailment algorithm is proposed to match the question templates,and word similarity algorithm based on Word Net and the semantic technology of dependency analysis are used to improve template matching algorithm based on the textual entailment,so that the precision,recall rate and F1 values of the question matching are improved.This paper mainly does the following five aspects of research work:(1)According to the relevant knowledge of SNOMED-CT,this paper makes full use of the semantic network formed by the concepts and their relationships in SNOMED-CT as the knowledge base of the biomedical question answering system to answer questions from physicians and ordinary users in natural language form,so as to realize the acquisition of relevant knowledge in the biomedical field.(2)Constructing a question template for multi-mapping structure based on SNOMED-CT and template logic.It makes the question templates can be mapped into the multiple semantic relationships in SNOMED-CT,so as to realize the diversity of inference rules.Through diversified inference rules,the semantics represented by the question template is closer to the user’s question intention.Thus,system can mostly satisfy user’s own question expectations.(3)A template description logic system is designed to tag template elements and define the structure of templates so as to accurately represent question semantics.Thus improving the processing efficiency of user questions,and laying the foundation for the accurate match of question template.The template logic system takes the biomedical ontology element as the operation object,and is a special description logic for semantic annotation and semantic interpretation of the question template.(4)A semantic-based textual entailment algorithm is proposed to match the question template.Firstly,the benchmark textual entailment algorithm is adopted to template match.Then,the ontology elements,word similarity algorithm in Word Net and dependency are added respectively to improve matching algorithm.In the process of algorithm improvement,the F1 values obtained by different matching thresholds are compared and analyzed by experiment,and the thresholds under the optimal F1 value are adopted in template match to further improve the flexibility and accuracy of the system.(5)The inference function bound to the question template is designed.Firstly,and the related information of the SNOMED-CT ontology element in the user question is obtained by the corresponding inference rule as a parameter and passed into the inference function;According to the information of the passed ontology element,the inference function performs the query in the SNOMED-CT ontology and operates on the set related to the ontology element to realize the answer reasoning.The experimental results show that the performance of the system has reached a high level when using the multiple mapping templates and the proposed algorithm,which can be used as a biomedical question answering system in the actual environment.