Print Email Facebook Twitter Supporting users with credibility assessments of health Tweets Title Supporting users with credibility assessments of health Tweets Author Post, F. Contributor Gao, Q. (mentor) Faculty Electrical Engineering, Mathematics and Computer Science Department Web Information Systems Date 2014-08-26 Abstract Twitter is one of the most popular social media platforms nowadays with more than 500 million tweets being send on a daily basis. A great variety of topics is discussed on Twitter and health has become a widely mentioned topic on Twitter. Users share updates regarding their own personal health situation and they use Twitter to look up health information. Inexperienced users have difficulty assessing the credibility of information contained in tweets. Additional visual and textual cues can be added to the tweets that enhance a users ability to determine the credibility. We define different types of feature sets to represent each tweet, which are then extracted and these features are used for automatic methods that can determine the credibility of health tweets. We narrow our domain to tweets about cancer and for each tweet we assess whether it’s credible, neutral or non-credible. We also investigate the different types of cancer tweets and how the credibility differs between the categories of tweets. We evaluate our research by using a labeled ground truth obtained via the crowd-sourcing platform CrowdFlower. Our results show that automatically supporting users with credibility assessments of health tweets is feasible and can be employed in practice but still leaves room for improvement. Subject twitterhealthclassifiercredibilitycrowdsourcing To reference this document use: http://resolver.tudelft.nl/uuid:ade2482a-15fc-42ba-bb1e-a5b6d9dfa842 Part of collection Student theses Document type master thesis Rights (c) 2014 Post, F. Files PDF thesis.pdf 799.78 KB Close viewer /islandora/object/uuid:ade2482a-15fc-42ba-bb1e-a5b6d9dfa842/datastream/OBJ/view