#677: From Words to Networks: AI-Powered Knowledge Graphs for Understanding Collaborative Learning

This paper presents an analytics tool that leverages knowledge graphs to make visible the complex, collaborative discourse in Knowledge Forum. The tool extracts relational triplets (subject, predicate, object) from student posts using a hybrid pipeline combining LLM-based, transformer-based (Triplex), and rule-based (spaCy) methods. Extracted triplets are normalized and visualized as interactive graphs, where nodes represent concepts and edges represent relationships. A case study on the salmon life cycle demonstrates how the visualization surfaces central ideas, thematic clusters, and knowledge gaps, aligning with key Knowledge Building principles. By transforming unstructured discourse into an explorable network, the tool supports students and educators in tracing idea development and identifying opportunities for deeper inquiry.