Diabetes-related topic detection in Chinese health websites using deep learning

Xinhuan Chen, Yong Zhang, Chunxiao Xing, Xiao Liu, Hsinchun Chen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Scopus citations


With 98.4 million people diagnosed with diabetes in China, most of the Chinese health websites provide diabetes related news and articles in diabetes subsection for patients. However, most of the articles are uncategorized and without a clear topic or theme, resulting in time consuming information seeking experience. To address this issue, we propose an advanced deep learning approach to detect topics for diabetes related articles from health websites. Our research framework for topic detection on diabetes related articles in Chinese is the first one to incorporate deep learning in topic detection in Chinese. It can identify topics of diabetes articles with high performance and potentially assist health information seeking. To evaluate our framework, experiment is conducted on a test bed of 12,000 articles. The results showed the framework achieved an accuracy of 70% in detecting topics and significantly outperformed the SVM based approach.

Original languageEnglish (US)
Title of host publicationSmart Health - International Conference, ICSH 2014, Proceedings
Number of pages12
ISBN (Print)9783319084152
StatePublished - 2014
Event2nd International Conference for Smart Health, CSH 2014 - Beijing, China
Duration: Jul 10 2014Jul 11 2014

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8549 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Other2nd International Conference for Smart Health, CSH 2014


  • Chinese
  • classification
  • deep learning
  • diabetes
  • topic detection

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Diabetes-related topic detection in Chinese health websites using deep learning'. Together they form a unique fingerprint.

Cite this