Modeling and Visualization of Group Knowledge Construction based on Cohesion Metrics in Data Inquiry Learning

Abstract

Group knowledge construction is seen as a symbol of effective collaboration. The quality of collaborative knowledge construction can be understood through the extended analysis of discourse. Cohesion is the basis of dialogue discourse, indicating the consistency of contextual topics in conversation. The study adopts natural language processing (NLP) and machine learning approaches based on discourse cohesion metrics to model and visualize the process of group knowledge construction. The three dimensions of cohesion metrics includes internal cohesion, social impact and responsivity. A group conversation dataset (participant N = 3, utterance N = 2,595) in the context of data inquiry learning is used for analyzing individual performance. Combined with the analysis of the actual conversation content, the visualization results show that it can describe the performance of participants in the group knowledge construction effectively. It has great potential to assist instructors to monitor and evaluate each participant’s performance in group discussion efficiently and provide guidance and scaffolds from the perspective of collaboration quality.