Abstract

Epidemic outbreaks detection and monitoring is an important but challenging task in epidemic control strategies. In recent years social media has been seen as a promising data source to track epidemic disease. Epidemic detection approaches often rely on data mined from Twitter and Facebook. These data can be geolocated in two ways: either based on geographic coordinates of the location from where a tweet or a post has been submitted if available, or based on place names mentioned in the text posted. In this thesis I propose to further explore the potential of place names in tweets to track a specific disease outbreak: The 2016 Zika outbreak. To explore this potential I have first collected about 1 million of tweets mentioning “Zika” during a period of 15 weeks. I have then geoparsed this database using different approaches to identify Twitter activity related to Zika for 13 selected countries. I have systematically compared these results with the official number of new cases of Zika recorded by official organizations for each country, every week. The results of this first set of analysis show that the degree of correlation between the volume of tweets and the number of official new cases is overall pretty low. Throughout this analysis, I was able to identify that the volume of Zika related tweets was largely affected by events not directly related to this disease (e.g. Venezuela struggles to contain Zika outbreak amid economic crisis). In order to better understand the nature and the impact of these events, I have done an in-depth qualitative analysis focusing on one case study: Venezuela. Although for Venezuela the quantitative analysis showed a strong correlation between the number of tweets and the number of new official Zika cases per week, the qualitative content analysis confirmed that very few of these Zika related tweets talk about new cases. In fact I was not able to identify even one Twitter account that would consistently provide information about new Zika cases while the disease was spreading out throughout the country. Based on these results I was able to emphasize the inappropriateness of using Twitter alone to try to track the spread of a disease, as well as the extensive use of Zika as a keyword for a large number of individuals and organizations to push other political and economic agendas.