Content Adaptation for Language Learning: A Hybrid AI Approach

In learning a foreign language, access to comprehensible input is a critical success factor. However, at early stages, when learners are still below an intermediate-proficiency level, finding level-appropriate and engaging materials is highly problematic. Although the Internet abounds in text and multimedia materials in many languages, most of them are too difficult to be useful for lower-proficiency language learners. The present project aimed to establish whether the affordances of large language models (LLMs) can be harnessed to turn authentic audio, video, and text materials into comprehensible input for independent elementary-level language learners. The present article reports on the outcomes of a research and development project that adopts a hybrid approach to simplifying authentic materials, combining affordances of LLMs with careful prompt engineering and rule-based refinement. The article details the hybrid sequential pipeline system and the results of two rounds of evaluation: language teacher ratings and automated text analysis indices. Based on the outcome of these evaluations, it is concluded that the proposed approach can provide an efficient way of simplifying authentic content for and by lower-proficiency language learners. Directions for future research and development are also proposed.