Cultural and geolocation aspects of communication in Twitter

Web applications exploit user information from social networks and online user activities to facilitate interaction and create an enhanced user experience. Due to privacy issues, however, it might be difficult to extract user data from social network, in particular location data. For instance, information on user location depends on users’ agreement to share own geographic data. Instead of directly collecting personal user information, we aim to infer user preferences by detecting behaviour patterns from publicly available microblogging content and users’ followers’ network. With statistical and machine-learning methods, we employ Twitter-specific features to predict country origin of users on Twitter with an accuracy of more than 90% for users from the most active countries. We further investigate users’ interpersonal communication with their followers. Our findings reveal that belonging to a particular cultural group is playing an important role in increasing users responses to their friends. The knowledge on user cultural origins thus could provide a differentiated state-of-the-art user experience in microblogs, for instance, in friend recommendation scenario.