Lyrica: A Web-Based Framework for Foreign Language Learning Through Music and Artificial Intelligence

Abstract

Research indicates that music enhances language learning by increasing engagement, motivation, and self-confidence while reducing anxiety [1]. However, most language learning applications do not integrate music, and those that do often offer limited functionality and lack user-friendly interfaces. Thus, this thesis designs, implements and evaluates Lyrica, a web-based framework that incorporates music and large language models to create an engaging language learning experience. The goal of Lyrica is to help learners improve their vocabulary, comprehension, and conversational skills in the target language by listening, studying, and discussing songs. The framework provides three types of assessments to test the progress of users in the aforementioned skills. The vocabulary assessment is conducted through text-based question-and-answer (Q&A) interactions and focuses on evaluating a user’s understanding of key words and phrases in a song. The dialogue assessment is an interactive conversation with a large language model, conducted either through text or speech, primarily designed to assess a user’s pronunciation and listening skills, while also engaging other aspects of language proficiency. Lastly, the comprehension assessment consists of multiple-choice questions that examine a user’s understanding of a song’s content and central themes. The initial evaluation of Lyrica highlights its strong potential to enhance foreign language learning through the integration of music and large language models.