Speakers
Description
This study explores the effectiveness of AI-generated songs in remedying fossilized errors, persistent pronunciation errors that challenge adult English language learners (N=46) despite extended exposure to the target language. Utilizing a quasi-experimental design, the study evaluates the impact of AI-developed songs tailored to address specific phonemic deficiencies. Participants who have developed fossilized errors are allocated into experimental and control groups. The experimental cohort undergoes instruction augmented by AI-generated songs, while the control group adheres to conventional pronunciation teaching methodologies. Pronunciation improvement is assessed through pre- and post-tests, concentrating on the accuracy of targeted phonemes.
The findings demonstrate a significant reduction in fossilized errors and a heightened proficiency in articulating individual sounds among participants exposed to AI-generated songs, compared to the control group. Furthermore, the study investigates the influence of individual preferences for songs and singing on the efficacy of this pedagogical approach. Qualitative insights gleaned from semi-structured learner interviews shed light on the motivational impact of AI-generated songs on pronunciation practice, underscoring their potential as a catalyst for pedagogical innovation. Ultimately, these findings contribute to the refinement of pronunciation instruction methodologies for adult learners, emphasizing the instrumental role of AI technology in language education.
