Yang Hu1, Guihua Wen2, Jiajiong Ma3, Danyang Li4, Changjun Wang5, Huihui Li6, Eryang Huan7. 1. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510000, China. Electronic address: superhy199148@hotmail.com. 2. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510000, China. Electronic address: crghwen@scut.edu.cn. 3. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510000, China. Electronic address: mullma@outlook.com. 4. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510000, China. Electronic address: danyangedu@163.com. 5. Guangdong General Hospital, Guangzhou 510000, China. Electronic address: gzwchj@126.com. 6. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510000, China. Electronic address: 29777562@qq.com. 7. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510000, China. Electronic address: huaneryang@gmail.com.
Abstract
BACKGROUND: Current Chinese medicine has an urgent demand for convenient medical services. When facing a large number of patients, understanding patients' questions automatically and precisely is useful. Different from the high professional medical text, patients' questions contain only a small amount of descriptions regarding the symptoms, and the questions are slightly professional and colloquial. OBJECT: The aim of this paper is to implement a department classification system for patient questions. Patients' questions will be classified into 11 departments, such as surgery and others. METHODS: This paper presents a morpheme growth model that enhances the memories of key elements in questions, and later extracts the "label-indicators" and germinates the expansion vectors around them. Finally, the model inputs the expansion vectors into a neural network to assign department labels for patients' questions. RESULTS: All compared methods are validated by experiments on three datasets that are composed of real patient questions. The proposed method has some ability to improve the performance of the classification. CONCLUSIONS: The proposed method is effective for the departments classification of patients questions and serves as a useful system for the automatic understanding of patient questions.
BACKGROUND: Current Chinese medicine has an urgent demand for convenient medical services. When facing a large number of patients, understanding patients' questions automatically and precisely is useful. Different from the high professional medical text, patients' questions contain only a small amount of descriptions regarding the symptoms, and the questions are slightly professional and colloquial. OBJECT: The aim of this paper is to implement a department classification system for patient questions. Patients' questions will be classified into 11 departments, such as surgery and others. METHODS: This paper presents a morpheme growth model that enhances the memories of key elements in questions, and later extracts the "label-indicators" and germinates the expansion vectors around them. Finally, the model inputs the expansion vectors into a neural network to assign department labels for patients' questions. RESULTS: All compared methods are validated by experiments on three datasets that are composed of real patient questions. The proposed method has some ability to improve the performance of the classification. CONCLUSIONS: The proposed method is effective for the departments classification of patients questions and serves as a useful system for the automatic understanding of patient questions.