Using content and text classification methods to characterize team performance

Because of the critical role that communication plays in a team's ability to coordinate action, the measurement and analysis of online transcripts in order to predict team performance is becoming increasingly important in domains such as global software development. Current approaches rely on human experts to classify and compare groups according to some prescribed categories, resulting in a laborious and error-prone process. To address some of these issues, the authors compared and evaluated two methods for analyzing content generated by student groups engaged in a software development project. A content analysis and semi-automated text classification methods were applied to the communication data from a global software student project involving students from the US, Panama, and Turkey. Both methods were evaluated in terms of the ability to predict team performance. Application of the communication analysis' methods revealed that high performing teams develop consistent patterns of communicating which can be contrasted to lower performing teams.

5th IEEE Conference on Global Software Engineering Princeton, USA, August 23-26, 2010.