Research on Text-based Q&A System Technology for Medical Aesthetic Field
JunZhen Liu, Yixuan Zhang, Bin Gong, Wei Lu, Yue Xiao · 2022
Medical aesthetics is the process of re-homing the shape of various parts of the body using medical means in order to meet the subjective definition of beauty. A text-based question and answer system in the field of medical aesthetics can provide quick answers to patients’ various medical search needs and help them to consult online and get professional explanations of their medical aesthetic problems. We propose the GPT2 Medchat model as an important part of the system and need to set up the following aspects to complete the system: 1. We prepare a new dataset in the medical aesthetics domain, which covers a wide range of questions and answers with strong professionalism. 2. In the pre-training phase, high-quality dialogue texts in the medical aesthetics domain are used as training materials. The system is built based on the pre-training phase. The medical text is used as training material, and the GPT2 Medchat model is trained and fine-tuned by relying on the pkuseg professional medical model to segment the medical text. The optimized segmentation results are used as a medical dictionary. 3. Sampling strategy to ensure the quality and diversity of the quizzes under the specified medical aesthetic domain questions. Both objective and subjective evaluations show that the GPT2 Medchat model has achieved a high practical and reference value for quizzing in the medical aesthetics domain.