Domain Supervised Deep Learning Framework for Detecting Chinese Diabetes-Related Topics

As millions of people are diagnosed with diabetes every year in China, many diabetes-related websites in Chinese provide news and articles. However, most of the online articles are uncategorized or lack a clear or unified topic, users often cannot find their topics of interest effectively and efficiently. The problem of health text classification on Chinese websites cannot be easily addressed by applying existing approaches, which have been used for English documents, in a straightforward manner. To address this problem and meet users’ demand for diabetes-related information needs, we propose a Chinese domain lexicon, adopt some professional diabetes topic explanations as domain knowledge and incorporate them into deep learning approach to form our topic classification framework. Our experiments using real datasets showed that the framework significantly achieved a higher effectiveness and accuracy in categorizing diabetes-related topics than most of the state-of-the-art benchmark approaches. Our experimental analysis also revealed that some health websites provided some incorrect or misleading category information.