StudyBuddy: An AI-Powered RAG Learning Ecosystem with Emotion-Aware Adaptive Hypothesis Generation

This paper presented StudyBuddy, an AI-powered learning ecosystem that integrates three complementary intelligent sub-systems into a unified educational platform: a FAISS-backed RAG engine for document-grounded question answering, a Random Forest emotion classifier trained on Kaggle text data, and a Find-S algorithm hypothesis learner that builds and forwards a continuously updated student knowledge profile to the LLM. Together, these components enable emotionally calibrated, personalised, and factually grounded responses that adapt dynamically to each student’s affective state and learning trajectory. Experimental results confirm strong emotion classification (macro-F1: 0.83), high RAG relevance (87% user rating), and a clear user preference for hypothesis-augmented responses (74% of paired comparisons). The voice interface achieves a 6.2% WER and 1.8-second latency, making speech interaction practical in everyday study sessions. Future work includes: semantic topic modelling to replace the keyword-based topic clustering in the hypothesis feature space; adaptive difficulty adjustment in flashcard generation based on the emotion and hypothesis state; multi-modal in-put (images, lecture slides); federated learning to improve the emotion model across institutions while preserving data privacy; and integration of an intelligent scheduling module to recommend study times based on the learner’s emotional trajectory across sessions.

[1]     P. Lewis, E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, H. Ku¨ttler, M. Lewis, W.-T. Yih, T. Rockta¨schel, S. Riedel, and D. Kiela, “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” in Proc. Advances in Neural Information Processing Systems (NeurIPS), vol. 33, pp. 9459–9474, 2020. [2]     M. Douze, A. Guzhva, C. Deng, J. Johnson, G. Szilvasy, P.-E. Mazare´, M. Lomeli, L. Hosseini, and H. Je´gou, “The Faiss Library,” IEEE Transactions on Big Data, 2024. DOI: 10.1109/TBDATA.2024.3348836. [3]     K. Macho´va´, M. Szabo´ova´, J. Paralicˇ, and M. Micˇko, “Detection of Emotion by Text Analysis Using Machine Learning,” Frontiers in Psychology, vol. 14, 2023. DOI: 10.3389/fpsyg.2023.1190326. [4]     Y. Zhao, J. Wang, Q. Li, and X. Chen, “Retrieval-Augmented Generation for Educational Application: A Systematic Survey,” Computers and Edu-cation: Artificial Intelligence, 2025. DOI: 10.1016/j.caeai.2025.100578. [5]     M. Klesel and H. F. Wittmann, “Retrieval-Augmented Generation (RAG),” Business & Information Systems Engineering, vol. 67, no. 4, pp. 551–561, 2025. [6]     I. Gligorea, M. Cioca, R. Oancea, A.-T. Gorski, H. Gorski, and P. Tudo-rache, “Adaptive Learning Using Artificial Intelligence in e-Learning: A Literature Review,” Education Sciences, vol. 13, no. 12, p. 1216, 2023. DOI: 10.3390/educsci13121216. [7]     M. Alshammari, R. Alshammari, and M. Hooper, “Personalised Adaptive Learning Technologies Based on Machine Learning Techniques to Identify Learning Styles: A Systematic Literature Review,” IEEE Access, vol. 11, 2023. DOI: 10.1109/ACCESS.2023.3276439. [8]     A. Neumann, Y. Yin, S. Sowe, S. Decker, and M. Jarke, “An LLM-Driven Chatbot in Higher Education for Databases and Information Systems,” IEEE Transactions on Education, vol. 68, pp. 103–116, 2025. DOI: 10.1109/TE.2024.3469120. [9]     F. A. Acheampong, H. Nunoo-Mensah, and W. Chen, “A Review on Emotion Detection by Using Deep Learning Techniques,” Artificial Intelligence Review, 2024. DOI: 10.1007/s10462-024-10831-1. [10]    M. Riyadh, M. Li, F. H. Lie, J. L. Loh, H. Mi, and S. Bohra, “LLM-Assisted Vector Similarity Search,” arXiv preprint arXiv:2412.18819, 2024.