Automated Team Discourse Modeling: Test of Performance and Generalization
Ahmed Abdelalí, Peter W. Foltz, Melanie J. Martin, Rob Oberbreckling, Mark Rosenstein · eScholarship (California Digital Library) · 2006
Team communication provides a rich source of discourse, which can be analyzed and tied to measures of team performance.Our goal is to better understand and model the relationship between team communication and team performance to improve team process, develop collaboration aids, and improve the training of teams.In the present work, we use Latent Semantic Analysis (LSA) for automating the analysis and annotation of team discourse.We describe two approaches to modeling team performance.The first measures the semantic content of a team's dialogue as a whole to predict the team's performance.The second categorizes each team member's statements using an established set of discourse tags and uses them to predict team performance.In three experimental settings we demonstrate the ability of these approaches to model performance and generalize to new datasets.