Lecture comprehension remains a critical bottleneck in higher education. Students retain less than 20% of auditory information after two weeks, and reviewing hours of raw audio is prohibitively time-consuming. This paper presents LECTRA, an AI-based Lecture Assistant System integrating four core NLP capabilities: Automatic Speech Recognition (ASR) using OpenAI Whisper, abstractive text summarisation using Facebook BART-Large-CNN, a Retrieval-Augmented Generation (RAG) question-and-answer module using RoBERTa, and automated quiz generation — all unified within a FastAPI web platform backed by SQLite. The system accepts audio, video, PDF, and plain-text inputs as well as YouTube URLs. Evaluation on 40 authentic academic lecture recordings (~20 hours) at Babcock University demonstrated a Word Error Rate (WER) of 9.8% on clean audio, ROUGE-1 of 0.4231, ROUGE-2 of 0.1987, ROUGE-L of 0.3814, and BLEU of 0.2156 for summarisation. A System Usability Scale (SUS) study with 12 undergraduate participants yielded 76.3/100 (‘Good’). LECTRA handles varied regional accents without subscription requirements, offering a practical, low-cost tool that advances educational NLP.
This paper presented LECTRA, an AI-based Lecture Assistant System that combines Automatic Speech Recognition using OpenAI Whisper, abstractive text summarisation using Facebook BART-Large-CNN, RAG-based question-answering using RoBERTa, and automated quiz generation — all integrated within a single FastAPI web platform. The system was built and evaluated on 40 authentic academic lecture recordings at Babcock University, Nigeria.
Empirical evaluation demonstrated reliable transcription accuracy (WER: 9.8% on clean audio), competitive summarisation performance (ROUGE-1: 0.4231; 7.4x compression ratio), and good usability (SUS: 76.3/100). All five project objectives were achieved. LECTRA contributes a practical, reproducible pipeline demonstrating how multiple pre-trained NLP models can be orchestrated for academic language processing in resource-constrained, subscription-free deployments.
Future work will explore: integration of Whisper Large-v2 for improved accuracy on noisy and accented audio; fine-tuning BART on a domain-specific corpus of academic lecture summaries; multilingual support for Yoruba, Hausa, and French to serve Nigeria’s linguistically diverse student population; mobile application development; and large-scale user trials across multiple institutions.
References
[1] D. A. Bligh, What’s the Use of Lectures? San Francisco: Jossey-Bass, 2000.
[2] R. Owston, D. Lupshenyuk, and H. Wideman, “Lecture capture in large undergraduate classes: Student perceptions and academic performance,” The Internet and Higher Education, vol. 14, no. 4, pp. 262–268, 2011.
[3] D. Jurafsky and J. H. Martin, Speech and Language Processing, 3rd ed. (draft). Stanford University, 2023. [Online]. Available: https://web.stanford.edu/~jurafsky/slp3/
[4] W. Holmes, M. Bialik, and C. Fadel, Artificial Intelligence in Education: Promises and Implications for Teaching and Learning. Boston: Center for Curriculum Redesign, 2019.
[5] K. A. Kiewra, “Investigating notetaking and review: A depth of processing alternative,” Educational Psychologist, vol. 20, no. 1, pp. 23–32, 1985.
[6] A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervision,” in Proc. Int. Conf. Machine Learning (ICML), 2023.
[7] M. Lewis et al., “BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension,” in Proc. 58th Annual Meeting of the Association for Computational Linguistics (ACL), 2020, pp. 7871–7880.
[8] P. Lewis et al., “Retrieval-augmented generation for knowledge-intensive NLP tasks,” in Proc. Advances in Neural Information Processing Systems (NeurIPS), vol. 33, 2020, pp. 9459–9474.
[9] Y. Liu et al., “RoBERTa: A robustly optimized BERT pretraining approach,” arXiv preprint arXiv:1907.11692, 2019.
[10] A. Bangor, P. Kortum, and J. Miller, “Determining what individual SUS scores mean: Adding an adjective rating scale,” Journal of Usability Studies, vol. 4, no. 3, pp. 114–123, 2009.
[11] S. Ahmed and M. Khan, “Smart tutor: NLP-based student query response system,” in Proc. Int. Conf. on Intelligent Systems, 2020, pp. 145–151.
[12] X. Li, Y. Wang, and Z. Chen, “Speech-to-text classroom assistant for accessibility in higher education,” Journal of Educational Technology, vol. 18, no. 2, pp. 77–89, 2021.
[13] Google Research, “AI note generator: Summarization algorithms for educational materials,” Technical Report, 2022.
[14] G. Murray, G. Carenini, R. Ng, and X. Liu, “Lecture summarization: A review of approaches,” Computational Linguistics, vol. 45, no. 2, pp. 345–385, 2019.
[15] T. Young, D. Hazarika, S. Poria, and E. Cambria, “Recent trends in deep learning based natural language processing,” IEEE Computational Intelligence Magazine, vol. 13, no. 3, pp. 55–75, 2018.
How to Cite This Paper
Joseph Oluwaseyi Adewuyi, Kehinde Gladys Akande, Victory Winifred Ebu-iduze, Joshua Ajisafe (2026). LECTRA: An AI-Based Lecture Assistant System Integrating Automatic Speech Recognition, Abstractive Summarisation, and Retrieval-Augmented Q&A. International Journal of Computer Techniques, 13(3). ISSN: 2394-2231.
