Multilingual Automatic Lyric Transcription

Multilingual automatic lyric transcription (ALT) aims to automatically convert song audio into written lyrics across multiple languages, a challenging task due to the complexities of singing voice and limited multilingual datasets. Current research focuses on adapting speech recognition models, such as wav2vec 2.0, to the ALT task, exploring both monolingual and multilingual approaches, and incorporating multimodal data (audio, video, IMU) to improve robustness. These advancements are significant for music information retrieval, enabling broader access to song lyrics for research and applications like music annotation, translation, and accessibility tools.