Topic categorization and representation of health community generated data

The representation and categorization of professional health provider released data have been well investigated and practically implemented. These have facilitated browsing, search and high-order learning of health information. On the other hand, there has been little corresponding studies on the representation and categorization of health community generated data. It is usually more complex, inconsistent and ambiguous, and consequently raises challenges for data access and analytics. This paper explores various representations for health community generated data and categorizes these data in terms of health topics. In addition, this work utilizes pseudo-labeled data to train the supervised topic categorization models, and this makes the whole categorization process unsupervised and extendable to handle large-scale data. The extensive experiments on two real-world datasets reveal our interesting findings of the informative representation approaches and effective categorization models for health community generated data.