HYBRID INTENT RECOGNITION FOR PERSONALIZED LEARNING: INTEGRATING DEEP LEARNING AND RULE-BASED MODELS IN MOODLE CHATBOTS

Abstract

Conversational chatbots have revolutionized educational platforms by providing personalized support and interactive learning experiences. However, existing intent recognition models in Learning Management Systems (LMS) like Moodle struggle with ambiguous, context-dependent queries, leading to inaccuracies in understanding student interactions. This research proposes a hybrid intent recognition algorithm that combines Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) deep learning models with rule-based verification to enhance the chatbot's ability to classify intents with higher precision. The primary objective of this study is to address the limitations of traditional Natural Language Understanding (NLU) techniques by leveraging deep learning for context retention and rule-based mechanisms for domain- specific accuracy. The proposed hybrid model was trained using a dataset of student queries collected from Moodle logs, categorized into 20 intent classes, and processed using tokenization, text cleaning, and word embeddings (e.g., GloVe, BERT). Comparative performance analysis against standalone models—Rule-Based, Naïve Bayes, LSTM, and GRU—demonstrated that the hybrid model achieved 92.5% accuracy, significantly outperforming conventional approaches. Statistical validation using paired t-tests and ANOVA confirmed the statistical significance of the improvements, with the hybrid model achieving the highest precision, recall, and F1- score while reducing response time by 20% compared to LSTM. These findings validate the proposed approach as a robust solution for real-time intent recognition in educational chatbots, fostering personalized learning, improving student engagement, and optimizing Moodle’s teaching-learning evaluation processes. Future research will explore reinforcement learning techniques to dynamically adapt chatbot responses and expand multilingual support for diverse educational environments.