Computational methods for mining health communications in web 2.0

摘要: Data from social media platforms are being actively mined for trends and patterns of interests. Problems such as sentiment analysis prediction election outcomes have become tremendously popular due to the unprecedented availability interactivity data different types. In this thesis we address two problems that been relatively unexplored. The first problem relates mining beliefs, in particular health their surveillance using media. second investigation factors associated with engagement U.S. Federal Health Agencies via Twitter Facebook. addressing propose a novel computational framework belief surveillance. This can be used 1) any given form probe, 2) automatically harvesting health-related probes. We present our estimates support, opposition doubt these probes some which represent true information, sense they supported by scientific evidence, others false information remaining debatable propositions. show example levels support surprisingly high. also study novelty find harvested sparse evidence may indicate hypothesis. suitability off-the-shelf classifiers quite generalizable classifying newly Finally, ability