A Question and Answering System for Mental Health of the Elderly Based on BiLSTM-CRF Model and Knowledge Graph

Currently, the aging population in China is becoming increasingly severe. Research has shown that 85% of the elderly have varying degrees of psychological problems, and 27% of the elderly have obvious psychological disorders such as anxiety and depression. The demand for offline and online consultation and intervention services is becoming increasingly urgent. However, there is currently a lack of systematic mental health intervention strategies targeting the elderly in relevant research and application practices. In order to explore new intervention services such as online consultation and chat with the elderly on mental health, a prototype system of question and answering for the elderly mental health has been designed, focusing on providing the elderly and related caregivers with daily psychological counseling services for the elderly. Firstly, a public mental health Q&A dataset has been collected, and then the data have been manually screened to obtain mental health Q&A dataset for the elderly. Based on this dataset, combined with the semantic data related to neurology for the elderly, the knowledge graph of mental health of the elderly has been constructed. Then, the problem matching module has been used to determine and analyze the types of problems faced by the elderly and generate corresponding answers to the questions. For the questions related to psychological counseling of the elderly, the BERT and BiLSTM-CRF network have been used to calculate the question template closest to the user’s question in the system, and find the corresponding answer matching the question in the knowledge graph. The experimental results show that the system can effectively understand the intention of elderly users to ask questions, and has good accuracy and reliability in answering elderly mental health related questions, which helps to address the mental health service needs of the elderly.