Research On The Construction Of Ethnic Medicine Knowledge Graph And Question-answering Syste

Ethnic medicine,which has endured for thousands of years,not only possesses unique treatment techniques and complete medical theories,but also fully embodies its regional,ethnic and inherited characteristics,and is an indispensable part of the excellent Chinese culture.The development of ethnic medicine has been seriously hampered by such practical factors as the remoteness of the region and the destruction of the herbal environment.With the help of Internet technology to mine and organize ethnomedicine data,we offer users intelligent information knowledge services in the area of ethnic medicine and boost the development and transmission of ethnic medicine.Therefore,thesis establishes a knowledge graph of ethnic medicine and designs an ethnic medicine question answering system using techniques such as interrogative similarity,named entity recognition and intention classification.The main work of this paper is as follows:(1)Based on ethnic medicine knowledge graph construction.The main steps of knowledge graph construction are data processing,text annotation,knowledge extraction and knowledge storage,respectively.Firstly,we process ethnic medicine data,set entity labels according to text features and annotate the ethnic medicine corpus with entities.Ethnic medicine Named Entity Recognition is performed by ALBERT-BiLSTM-CRF model for training,and predefined relationships between entities,extracted triadic data and stored into Neo4j graph database for visualization of ethnic medicine knowledge graph.(2)Methodological design based on an ethnic medicine question answering system.Achieved in two ways.Firstly,the interrogative similarity matching is performed based on the Siamese network,and the experimental result F1 is 97%,which is set as the boundary value,if it is over the threshold,the answer is returned directly by retrieving the constructed question answering database.If it is less than the threshold,the question is pre-processed to identify the entities in the question,and the user intention is obtained by TextCNN,and the answer is queried in the knowledge graph by combining with the Cypher statement template.(3)Question and answer system implementation and testing.Based on the Django framework to implement the question and answer interface,the front end to display the user asked questions and answers returned,the back end to call the algorithm module and the database.At the same time,after testing the system to verify its reliability,it can satisfy the user’s question and answer needs and has high practical value.