The utility of social and topical factors in anticipating repliers in Twitter conversations
Publikation aus Digital
Schantl J., Kaiser R., Wagner C., Strohmaier M.
Paris, France Proceedings of the 5th Annual ACM Web Science Conference, p376-385, 2013
Anticipating repliers in online conversations is a fundamental challenge for computer mediated communication systems which aim to make textual, audio and/or video communication as natural as face to face communication. The massive amounts of data that social media generates has facilitated the study of online conversations on a scale unimaginable a few years ago. In this work we use data from Twitter to explore the predictability of repliers, and investigate the factors which influence who will
reply to a message. Our results suggest that social factors, which describe the strength of relations between users, are more useful than topical factors. This indicates that Twitter users' reply behavior is more impacted by social relations than by topics. Finally, we show that a binary classification model, which differentiates between users who will and users who will not reply to a certain message, may achieve an F1-score of 0.74 when using social features.