Research on Automatic Discourse Classification during Collaborative Knowledge Construction: A Deep Learning Analysis Method Based on Semantic Extension

Analyzing collaborative knowledge construction is crucial for evaluating learning quality and instructional design. Collaborative learning, akin to dialogue, demands efficient methods for textual analysis. However, existing automatic analysis struggles with sparse discourse features, resulting in low accuracy. This paper introduces a semantic extension-based classification model for short collaborative learning texts, enhancing representational power through keywords and grammatical structures. Experimental evaluation on self-constructed data and Sina news headlines shows improved accuracy compared to baseline algorithms. Additionally, the model is applied to real collaborative learning activities, revealing variability in knowledge construction levels and behavior patterns among groups. Thus, it enables in-depth analysis and timely feedback on collaborative knowledge construction proficiency.