Finding transactive contributions in whole group classroom discussions

Automatic analysis of discussion logs from studies of computer supported collaborative learning has become more prevalent as technology for machine learning and text mining have become more powerful and usable. We extend this effort to analysis of transcripts of whole group classroom discussions. We argue in favor of adapting a transactivity based framework developed for analysis of collaborative learning discussions for use in the context of whole class discussions. Results from an evaluation of this technology applied to transcripts from 3 class periods of middle school math demonstrate promising results, specifically Cohen’s Kappa of .69 in comparison with human annotators for identifying transactive contributions and .68 for identifying which prior utterance a transactive contribution relates to. Potential applications for teacher professional development, automatic assessment of learning processes, and real time support for classroom discussion facilitators is discussed. Implications for theory building related to learning in classroom discussions is also addressed.