Automatic Generation of School Activities for Teaching English to Children

• Lorenna Marinho Lucena UFCG
• Matheus Lisboa Oliveira dos Santos UFCG
• Claudio E. C. Campelo UFCG

DOI: https://doi.org/10.5753/sbie.2025.12773

Resumo

Activity sheets are essential tools for developing basic skills, integrating theoretical and practical knowledge. For children, they should be playful, creative and interactive, facilitating the association of ideas. Creating these sheets is a challenge, requiring technical and pedagogical skills, especially in language teaching. This article proposes the use of Large Language Models (LLMs) with fine-tuning to generate English as a Second Language (ESL) exercises for children. Different models were evaluated and compared against fine-tuned GPT-3.5. The approach was evaluated with the help of ESL teaching experts, based on criteria such as naturalness, usefulness, diversity and personalization. The results indicated that the fine-tuning approach is promising.

Referências

Al Faraby, S., Adiwijaya, A., and Romadhony, A. (2024). Review on neural question generation for education purposes. International Journal of Artificial Intelligence in Education, 34(3):1008–1045.

Council, B. (2022). Diretrizes para o ensino de inglês nos anos iniciais. Acessado em: 31 mar. 2025.

da Silva, M. D. R. and do Rosário Sabota Silva, B. (2019). Perspectivas criticas de ensino de linguas: Produção e adaptação de materiais didativos. V CEPE, Universidade Estadual de Goiás, CCSEH.

Karpicke, J. D. and Roediger III, H. L. (2008). The critical importance of retrieval for learning. science, 319(5865):966–968.

Krashen, S. (1982). Principles and practice in second language acquisition.

Kriangchaivech, K. and Wangperawong, A. (2019). Question generation by transformers. arXiv preprint arXiv:1909.05017.

Kwon, T. (2023). Interfaces for personalized language learning with generative language models. PhD thesis, Columbia University.

Li, Z., Liu, H., Zhou, D., and Ma, T. (2024). Chain of thought empowers transformers to solve inherently serial problems. arXiv preprint arXiv:2402.12875, 1.

Neo, A. and Neo, G. (2024). Revisão sobre a geração automática de questões na educação: Técnicas, conjuntos de dados e métricas de avaliação.

Paivio, A. (2013). Imagery and verbal processes. Psychology Press.

Passos, C. and Silva, Z. (2014). Eu gosto m@is: 2º, 3º, 4º e 5º anos do Ensino Fundamental. IBEL, São Paulo, 2 edition. Coleção didática.

Rao, P. R., Jhawar, T. N., Kachave, Y. A., and Hirlekar, V. (2022). Generating qa from rule-based algorithms. In 2022 International Conference on Electronics and Renewable Systems (ICEARS), pages 1697–1703. IEEE.

Sonderegger, S. (2022). How generative language models can enhance interactive learning with social robots. International Association for Development of the Information Society.

Zelikman, E., Wu, Y., Mu, J., and Goodman, N. (2022). Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35:15476–15488.

Zhou, D., Schärli, N., Hou, L., Wei, J., Scales, N., Wang, X., Schuurmans, D., Cui, C., Bousquet, O., Le, Q., et al. (2022). Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625.