Identifying Multi-dimensional Information from Microblogs During Epidemics.

Microblogging sites like Twitter and Weibo have been shown to provide important real-time information during epidemics and disease outbreaks. During such situations, different types of stakeholders look for different types of information, such as symptoms, prevention, treatment schemes, death reports, and many more. Additionally, lots of personal opinions, sentiments are also posted on social media along with factual contents. In this work, we propose a method to automatically classify tweets posted during an epidemic into various informative categories. To this end, we utilize features derived from a medical knowledge base (UMLS) as well as features based on syntactic and lexical structure of tweets. We apply the classifier over tweets related to several diseases (Ebola, Dengue, and MERS), and show that, the proposed approach yields better classification performance as compared to earlier works. We also identify some interesting directions of future work, e.g., applying the classifier over drug addictions like Opioid.