Canal9: A database of political debates for analysis of social interactions

Automatic analysis of social interactions attracts major attention in the computing community, but relatively few benchmarks are available to researchers active in the domain. This paper presents a new, publicly available, corpus of political debates including not only raw data, but a rich set of socially relevant annotations such as turn-taking (who speaks when and how much), agreement and disagreement between participants, and role played by people involved in each debate. The collection includes 70 debates for a total of 43 hours and 10 minutes of material.