A dialogue system-driven teaching resource management platform for vocational education

This paper presents an intelligent teaching resource management platform built upon a task-oriented dialogue system, specifically tailored to the structured and domain-specific needs of vocational contexts. The system integrates a schema-guided dialogue state tracking framework that combines LSTM-based feature extraction, a slot gating mechanism, and a pointer neural network to effectively handle diverse conversational phenomena and unknown slot values. Experimental results show that the proposed model achieves a Joint Goal Accuracy (JGA) of 0.580, outperforming competitive baselines such as STAR (0.568) and TRADE (0.487). Unlike general-purpose chatbots, this platform supports real-time resource retrieval, pedagogical interaction, and administrative management in vocational settings, contributing to improved teaching quality and learner engagement.