oa Visualizing life in Qatar through the lens of twitter

In this interactive poster (a set of public web demos) we present insights into the socio-cultural activity in Qatar as seen on Twitter. Our starting point is the identification of around 70,000 active Twitter users that are, either permanently or temporarily, based in Qatar. This identification is done using a number of features such as (i) the self-stated location, (ii) the geo-tagged position of tweets from geo-enabled mobile devices, and (iii) a large amount of interest (= following) in local Twitter accounts such as @dohanews (“Doha News”), or @qatarevents (“I love Qatar”). Comparing to estimates from the Arab Social Media Report compiled by the Dubai School of Government we are confident that our sample includes the majority of active Twitter users in Qatar. For this user population we regularly obtain all their public tweets using the public Twitter API and annotate them with geographic co-ordinates (if the tweets comes from a mobile device) and with the language used. Language detection is performed using state-of-the-art libraries which work well for most cases but still fail for the comparatively small number of Tweets in Arabizi or other transliterated languages. The first demo visualizes the variation in language usage in different parts of Qatar in general and Doha in specific. Whereas English dominates in West Bay, Arabic dominates in less urban areas and Tagalog, the main Philippine language, is very prominent in the Industrial Area. This simple analysis already illustrates potential applications for more systematic social science studies. The second demo visualizes the “pulse of the city” and allows the user to select a day of the week, time of the day and language of interest. Areas that are more or less active than expected are highlighted which provides insights into where people live, work or are actives on weekends. For the last demo, we manually compiled topical lists of words related to topics such as traffic, sports, shopping, religious activities and references, family and friends, and work. These lists are translated to English, Arabic and Tagalog, the main languages in our data set. The user then has the option to see areas with an increased activity either in time (e.g. mosques for religion) or in space (e.g. Aspire for sports). We believe that this exploratory, interactive study shows promising potential for large scale demographic studies as well as for survey-type analysis to get insights into the worries and concerns of large parts of the populations. Though the Twitter population is clearly not a representative sample of the offline population, a bias could potentially be corrected by weighting different members in the sample differently. For example, older users or Indian workers, two groups that are underrepresented in our current data, could be given additional weight. We hope that our interactive presentation will initiate conversations and spark new ideas on how to use Social Media for social science research.