Testing a model of witness accounts in social media

Identifying micro-bloggers who are likely witnesses to events is beneficial in numerous applications, including event detection and credibility assessment. This paper presents research in-progress on testing of a conceptual model, which defines witness and related accounts from micro-blogs about events. The case study events considered have varying spatial and temporal characteristics, and include a shark sighting, a music concert, a protest, and a cyclone. Results indicate that witnessing characteristics are influenced by numerous factors in addition to the spatial and temporal characteristics of the events, including the motivation of the witnesses themselves. Additionally, the results suggest enhancements to the conceptual model to provide a more sophisticated generic implementation, and insights for future automation approaches.