Text, Topics & Sentiment

Statistical modelling of online text streams, annotator bias in sentiment analysis, and directional topic mixtures.

Online media and collaborative platforms generate rich text streams alongside structured user interactions. Our work in this area includes mixture models that correct for systematic bias across sentiment classifiers and human annotators, online trans-dimensional von Mises–Fisher directional mixtures for evolving user topic profiles, and network-text analyses of collaborative efficiency in online communities.

Current research in the group (funded through the AIMSIR Centre) extends social media sentiment and natural language modelling to real-time impact verification of severe weather warnings.

Project Publications

Qin, Xiangju, Pádraig Cunningham, and Michael Salter-Townshend. 2016. “Online Trans-Dimensional von Mises–Fisher Mixture Models for User Profiles.” Journal of Machine Learning Research 17 (200): 1–51.

Qin, Xiangju, Pádraig Cunningham, and Michael Salter-Townshend. 2015. “The Influence of Network Structures of Wikipedia Discussion Pages on the Efficiency of WikiProjects.” Social Networks 43: 1–15.

Qin, Xiangju, Michael Salter-Townshend, and Pádraig Cunningham. 2014. “Exploring the Relationship Between Membership Turnover and Productivity in Online Communities.” In Proceedings of the International AAAI Conference on Web and Social Media (ICWSM), 8:406–15. 1.

Salter-Townshend, Michael, and Thomas Brendan Murphy. 2014. “Mixtures of Biased Sentiment Analysers.” Advances in Data Analysis and Classification 8 (1): 85–103.

Salter-Townshend, Michael, and Thomas Brendan Murphy. 2012. “Sentiment Analysis of Online Media.” In Algorithms from and for Nature and Life. Studies in Classification, Data Analysis, and Knowledge Organization. Springer.