Mental Health Question and Answering System Based on Bert Model and Knowledge Graph Technology

Chaohui Guo,Shaofu Lin,Zhisheng Huang,Yahong Yao

DOI: https://doi.org/10.1145/3500931.3501011

2021-01-01

Abstract:With the development and progress of society, people are facing increasing pressure. The emergence of this phenomenon has led to a rapid increase in the incidence of mental illness. In order to deal with this phenomenon, this paper proposes a system of question and answering on the basic knowledge of mental health (MHQ&A) by using deep learning retrieval technology and knowledge graph technology. The system MHQ&A is designed mainly for the general public, to answer the basic knowledge of mental health, especially the field of depression. First of all, the basic and the professional question and answer data about mental health were respectively obtained by the reptilian bot from the "IASK" website knowledge and the "Dr. Dingxiang" website. Then, the questions and answers obtained through the crawler are made into a Question and Answering Knowledge Graph of Basic Health Knowledge in the mental health field, which is combined with semantic data of antidepressants and the semantic data of depression papers. Finally, a set of template matching rules is designed to determine the type of problem of users. If the questions are about the professional knowledge of medicine or thesis, the reasoning template will be used to reason and search the answer in the "Question and Answering Knowledge Graph of Basic Health Knowledge in the Mental Health Field". If the questions are about other basic knowledge in the field of mental health, the BERT model is used to vectorize the questions of users, and the matching questions and corresponding answers in the MHQ&A are found through cosine similarity calculation. Through the test of system accuracy, it is proved that the system can effectively combine deep learning technology and knowledge.