LLMs for Easy Language Translation: A Case Study on German Public Authorities Web Pages

This paper examines the use of Large Language Models (LLMs) for the intralingual translation of documents from standard German to German Easy Language (Leichte Sprache). We use open-weight models, from the Llama 3 family, with less than ten billion parameters. Additionally, we employ parameter-efficient fine-tuning (QLoRA) to adapt the LLMs to the requirements of Easy Language. For this purpose, we introduce a new data set (ELGEPA) (Source code to obtain the dataset: github.com/minds-hh/ger-gov-easy-lang), which is a parallel corpus of governmental documents in standard German and Easy Language with additional metadata, obtained from all German federal states and their capitals In our experiments, a fine-tuned Llama 3.1-8B-Instruct model achieved a SARI score of 41 and Flesch Reading Ease of 69. Outperforming GPT-4o and indicating that this type of model can deliver promising text quality in Easy Language.

Recommended citation: Schomacker, T., Tinman, B., Biemann, C., Tropmann-Frick, M. (2026). LLMs for Easy Language Translation: A Case Study on German Public Authorities Web Pages. In: Braun, T., Paaßen, B., Stolzenburg, F. (eds) KI 2025: Advances in Artificial Intelligence. KI 2025. Lecture Notes in Computer Science(), vol 15956. Springer, Cham. https://doi.org/10.1007/978-3-032-02813-6_20
Download Paper | Download Manuscript