Understanding Students Through Dialogue: A Dialogue Knowledge Tracing System for Learning Analytics

摘要

While the emergence of large-scale language models has made educational chatbots much easier, enabling them to proactively guide conversations remains a challenge. To achieve this, the system must track users. To better track student progress, we propose an AI-assisted system that combines dialogue-based knowledge tracking with instructional content analysis, automatically inferring students' knowledge mastery from conversational interactions.To support proactive guidance, the system needs to track students' progress, therefore, in the third year, we propose an AI-assisted system that combines dialogue-based knowledge tracking with instructional content analysis, automatically inferring students' knowledge mastery from conversational interactions.Our system employs LightRAG to analyze instructional materials, generating knowledge graphs with conceptual contexts, and uses large language models to construct hierarchical mind maps as the foundation for knowledge tracing. We innovatively integrate knowledge tracing mechanisms into educational chatbot dialogues, enabling automatic annotation of students' knowledge point mastery and providing real-time tracking reports for personalized guidance. Visualization tools such as radar charts enable quick comparison of student performance across topics.We evaluated the system in a university Chinese literature course, comparing automated assessments with teachers' manual evaluations. Results show the system effectively reflects students' knowledge mastery with significant correlation to teacher judgments. Virtual student experiments demonstrate that real-time knowledge tracing enhances chatbot guidance effectiveness, particularly for passive learners. The system successfully implements "consolidating known knowledge" and "expanding new knowledge" teaching strategies, confirming personalized instructional guidance feasibility.This research integrates course material analysis, dialogue understanding, and learning trajectory visualization while innovatively applying knowledge tracing to real-time educational chatbot conversations, providing concrete references for conversational AI in smart education.