Abstract
We use large language models to aid learners enhance proficiency in a foreignlanguage. This is accomplished by identifying content on topics that the useris interested in, and that closely align with the learner's proficiency levelin that foreign language. Our work centers on French content, but our approachis readily transferable to other languages. Our solution offers severaldistinctive characteristics that differentiate it from existinglanguage-learning solutions, such as, a) the discovery of content across topicsthat the learner cares about, thus increasing motivation, b) a more preciseestimation of the linguistic difficulty of the content than traditionalreadability measures, and c) the availability of both textual and video-basedcontent. The linguistic complexity of video content is derived from the videocaptions. It is our aspiration that such technology will enable learners toremain engaged in the language-learning process by continuously adapting thetopics and the difficulty of the content to align with the learners' evolvinginterests and learning objectives.