Automatic Generation of Study Texts from Lecture Slides

• William Massami Costa Harada UEA
• Fabio Santos da Silva UEA

DOI: https://doi.org/10.5753/wesaac.2026.32082

Resumo

Generative artificial intelligence has expanded the potential for automated educational content generation to support teaching and independent learning. Among these resources, lecture slides play a central role, yet their concise visual format limits its use for asynchronous study by students. This work proposes a method to automatically transform slide decks into self-contained study documents. The solution integrates visual parsing, iterative discourse reconstruction, and knowledge retrieval to ground implicit lecture concepts. Experiments across four multimodal LLMs show the generation of coherent materials while highlighting operational trade-offs.

Referências

Bai, S., Cai, Y., Chen, R., Chen, K., Chen, X., Cheng, Z., Deng, L., Ding, W., Gao, C., Ge, C., et al. (2025). Qwen3-vl technical report. arXiv preprint arXiv:2511.21631.

Branch, R. M. and Varank, İ. (2009). Instructional design: The ADDIE approach, volume 722. Springer.

Chu, Z., Wang, S., Xie, J., Zhu, T., Yan, Y., Ye, J., Zhong, A., Hu, X., Liang, J., Yu, P. S., et al. (2025). Llm agents for education: Advances and applications. arXiv preprint arXiv:2503.11733, 2.

Google DeepMind (2025). Gemini 2.5 flash model card. [link]. Model Card.

Jacobs, S. and Jaschke, S. (2024). Leveraging lecture content for improved feedback: Explorations with gpt-4 and retrieval augmented generation. In 2024 36th International Conference on Software Engineering Education and Training (CSEE&T), pages 1–5. IEEE.

Liang, J., Stephens, J. M., and Brown, G. T. (2025). A systematic review of the early impact of artificial intelligence on higher education curriculum, instruction, and assessment. In Frontiers in Education, volume 10, page 1522841. Frontiers Media SA.

Moore, R. L. and Lee, S. S. (2024). Harnessing generative ai (genai) for automated feedback in higher education: A systematic review. Online Learning Journal.

Qwen Team (2026). Qwen3.7-plus. [link]. Accessed: 2026-06-20.

Rao, K., Coviello, G., Sankaradas, M., De Vita, C. G., Mellone, G., and Chakradhar, S. (2025). Slidecraft: Context-aware slides generation agent. In 2025 IEEE Conference on Pervasive and Intelligent Computing (PICom), pages 165–172. IEEE.

Swacha, J. and Gracel, M. (2025). Retrieval-augmented generation (rag) chatbots for education: A survey of applications. Applied Sciences, 15(8):4234.

Tanaka, R., Nishida, K., Nishida, K., Hasegawa, T., Saito, I., and Saito, K. (2023). Slide-vqa: A dataset for document visual question answering on multiple images. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 13636–13645.

Wang, Y., Yu, J., Zhang-Li, D., Lim, J. J. Y., Tu, S., Li, H., Liu, Z., Liu, H., Hou, L., Li, J., et al. (2025). Educraft: A system for generating pedagogical lecture scripts from long-context multimodal presentations. In Proceedings of the 34th ACM International Conference on Information and Knowledge Management, pages 6153–6160.

Wataoka, K., Takahashi, T., and Ri, R. (2024). Self-preference bias in llm-as-a-judge. arXiv preprint arXiv:2410.21819.

Wecker, C. (2012). Slide presentations as speech suppressors: When and why learners miss oral information. Computers & Education, 59(2):260–273.

Yao, H., Xu, W., Turnau, J., Kellam, N., and Wei, H. (2026). Instructional agents: Reducing teaching faculty workload through multi-agent instructional design. In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 4087–4109.

Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., and Cao, Y. (2022). React: Synergizing reasoning and acting in language models. arXiv preprint arXiv:2210.03629.