Developing a Conversational Chatbot with Word Embedding for Outpatient Clinic Appointments

• Main Authors: CHANG,YING-WEN, 張盈文
• Other Authors: YEH, HSIANG-YUAN
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/3wuf9h

碩士 === 東吳大學 === 巨量資料管理學院碩士學位學程 === 107 === In recent years, a great number of people begin to use hospital registration system through Internet technology. Although some hospitals have provided outpatient clinics and symptom comparison table for their reference, the combinations of symptoms and diseases are too complicated to make the proper appointment. The situation may also delay patient’s time and waste the medical cost. Therefore, we propose an outpatient clinic recommender system to assist patients in making appropriate department according to patients’ symptoms. At present, most of the previous studies established the correlations between diseases and symptoms via experts and the processes are time-consuming and labor-intensive. We first collect medical professional literature and Wikipedia content as our knowledge base. Second, we use Term Frequency - Inverse Document Frequency (TF-IDF) and the neural network based word2vec algorithm to automatically find the associations between diseases and symptoms. Finally, we apply network connection analysis method such as HITS and iterative random walk with restart methods and the concept of the Multi-armed Bandit theory to achieve the recommendation of the outpatient department through chatbot conversation process. The experimental results show that the combination of the word embedding techniques and iterative random walk with restart method gets the better results comparing with term frequency based method. Our approaches integrate graph theory and word embedding techniques for outpatient clinic recommendation and it can provide hospital for future planning reference.