Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
Current automatic lyrics transcription (ALT) benchmarks focus exclusively on word content and ignore the finer nuances of written lyrics including formatting and punctuation, which leads to a potential misalignment with the creative products of musicians and songwriters as well as listeners' experiences. For example, line breaks are important in conveying information about rhythm, emotional emphasis, rhyme, and high-level structure. To address this issue, we introduce Jam-ALT, a new lyrics transcription benchmark based on the JamendoLyrics dataset. Our contribution is twofold. Firstly, a complete revision of the transcripts, geared specifically towards ALT evaluation by following a newly created annotation guide that unifies the music industry's guidelines, covering aspects such as punctuation, line breaks, spelling, background vocals, and non-word sounds. Secondly, a suite of evaluation metrics designed, unlike the traditional word error rate, to capture such phenomena. We hope that the proposed benchmark contributes to the ALT task, enabling more precise and reliable assessments of transcription systems and enhancing the user experience in lyrics applications such as subtitle renderings for live captioning or karaoke.
PDF AbstractCode
Datasets
Results from the Paper
Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
---|---|---|---|---|---|---|
Automatic Lyrics Transcription | Jam-ALT | HTDemucs + Whisper v3 | Word Error Rate (WER) | 47.9 | # 5 | |
Case Error Rate | 3.8 | # 2 | ||||
Punctuation F1 | 29.0 | # 4 | ||||
Line break F1 | 65.7 | # 4 | ||||
Automatic Lyrics Transcription | Jam-ALT | AudioShake | Word Error Rate (WER) | 26.0 | # 1 | |
Case Error Rate | 3.4 | # 1 | ||||
Punctuation F1 | 50.5 | # 1 | ||||
Line break F1 | 82.3 | # 1 | ||||
Section break F1 | 72.1 | # 1 | ||||
Parenthesis F-1 | 29.4 | # 1 | ||||
Automatic Lyrics Transcription | Jam-ALT | Whisper v2 | Word Error Rate (WER) | 35.7 | # 3 | |
Case Error Rate | 4.5 | # 4 | ||||
Punctuation F1 | 41.7 | # 2 | ||||
Line break F1 | 69.3 | # 3 | ||||
Section break F1 | 3.3 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT | Whisper v3 | Word Error Rate (WER) | 35.5 | # 2 | |
Case Error Rate | 4.3 | # 3 | ||||
Punctuation F1 | 41.6 | # 3 | ||||
Line break F1 | 73.5 | # 2 | ||||
Section break F1 | 1.0 | # 3 | ||||
Automatic Lyrics Transcription | Jam-ALT | HTDemucs + Whisper v2 | Word Error Rate (WER) | 44.0 | # 4 | |
Case Error Rate | 5.3 | # 5 | ||||
Punctuation F1 | 28.0 | # 5 | ||||
Line break F1 | 61.2 | # 5 | ||||
Automatic Lyrics Transcription | Jam-ALT English | HTDemucs + Whisper v3 | Word Error Rate (WER) | 43.0 | # 5 | |
Case Error Rate | 4.1 | # 4 | ||||
Punctuation F-1 | 23.3 | # 6 | ||||
Line break F-1 | 66.8 | # 4 | ||||
Automatic Lyrics Transcription | Jam-ALT English | HTDemucs + Whisper v2 | Word Error Rate (WER) | 32.3 | # 3 | |
Case Error Rate | 5.3 | # 6 | ||||
Punctuation F-1 | 39.2 | # 3 | ||||
Line break F-1 | 53.8 | # 6 | ||||
Automatic Lyrics Transcription | Jam-ALT English | Whisper v2 | Word Error Rate (WER) | 43.8 | # 6 | |
Case Error Rate | 3.5 | # 2 | ||||
Punctuation F-1 | 31.3 | # 5 | ||||
Line break F-1 | 63.0 | # 5 | ||||
Section break F-1 | 11.2 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT English | LyricWhiz | Word Error Rate (WER) | 24.6 | # 2 | |
Case Error Rate | 3.5 | # 2 | ||||
Punctuation F-1 | 34.0 | # 4 | ||||
Line break F-1 | 74.0 | # 2 | ||||
Section break F-1 | 1.4 | # 4 | ||||
Automatic Lyrics Transcription | Jam-ALT English | AudioShake | Word Error Rate (WER) | 22.1 | # 1 | |
Case Error Rate | 3.4 | # 1 | ||||
Punctuation F-1 | 59.0 | # 1 | ||||
Parenthesis F-1 | 32.4 | # 1 | ||||
Line break F-1 | 80.7 | # 1 | ||||
Section break F-1 | 77.4 | # 1 | ||||
Automatic Lyrics Transcription | Jam-ALT English | Whisper v3 | Word Error Rate (WER) | 37.7 | # 4 | |
Case Error Rate | 4.8 | # 5 | ||||
Punctuation F-1 | 40.9 | # 2 | ||||
Line break F-1 | 71.5 | # 3 | ||||
Section break F-1 | 2.6 | # 3 | ||||
Automatic Lyrics Transcription | Jam-ALT French | Whisper v2 | Word Error Rate (WER) | 27.7 | # 1 | |
Case Error Rate | 3.2 | # 2 | ||||
Punctuation F-1 | 45.8 | # 1 | ||||
Line break F-1 | 73.4 | # 3 | ||||
Section break F-1 | 1.4 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT French | HTDemucs + Whisper v3 | Word Error Rate (WER) | 44.9 | # 5 | |
Case Error Rate | 3.2 | # 2 | ||||
Punctuation F-1 | 30.9 | # 5 | ||||
Line break F-1 | 69.4 | # 4 | ||||
Automatic Lyrics Transcription | Jam-ALT French | Whisper v3 | Word Error Rate (WER) | 34.7 | # 2 | |
Case Error Rate | 3.3 | # 5 | ||||
Punctuation F-1 | 42.4 | # 3 | ||||
Line break F-1 | 77.8 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT French | AudioShake | Word Error Rate (WER) | 34.9 | # 3 | |
Case Error Rate | 2.0 | # 1 | ||||
Punctuation F-1 | 45.8 | # 1 | ||||
Parenthesis F-1 | 41.3 | # 1 | ||||
Line break F-1 | 84.9 | # 1 | ||||
Section break F-1 | 72.5 | # 1 | ||||
Automatic Lyrics Transcription | Jam-ALT French | HTDemucs + Whisper v2 | Word Error Rate (WER) | 43.3 | # 4 | |
Case Error Rate | 3.2 | # 2 | ||||
Punctuation F-1 | 34.9 | # 4 | ||||
Line break F-1 | 66.1 | # 5 | ||||
Automatic Lyrics Transcription | Jam-ALT German | Whisper v2 | Word Error Rate (WER) | 45.4 | # 4 | |
Case Error Rate | 5.3 | # 4 | ||||
Punctuation F-1 | 38.7 | # 3 | ||||
Line break F-1 | 69.9 | # 4 | ||||
Automatic Lyrics Transcription | Jam-ALT German | AudioShake | Word Error Rate (WER) | 24.4 | # 1 | |
Case Error Rate | 4.1 | # 2 | ||||
Punctuation F-1 | 48.5 | # 1 | ||||
Parenthesis F-1 | 8.1 | # 1 | ||||
Line break F-1 | 81.2 | # 1 | ||||
Section break F-1 | 69.2 | # 1 | ||||
Automatic Lyrics Transcription | Jam-ALT German | Whisper v3 | Word Error Rate (WER) | 40.7 | # 2 | |
Case Error Rate | 4.0 | # 1 | ||||
Punctuation F-1 | 41.2 | # 2 | ||||
Line break F-1 | 71.2 | # 3 | ||||
Section break F-1 | 1.2 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT German | HTDemucs + Whisper v3 | Word Error Rate (WER) | 43.5 | # 3 | |
Case Error Rate | 4.4 | # 3 | ||||
Punctuation F-1 | 34.0 | # 4 | ||||
Line break F-1 | 72.0 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT German | HTDemucs + Whisper v2 | Word Error Rate (WER) | 65.2 | # 5 | |
Case Error Rate | 5.9 | # 5 | ||||
Punctuation F-1 | 30.2 | # 5 | ||||
Line break F-1 | 67.5 | # 5 | ||||
Automatic Lyrics Transcription | Jam-ALT Spanish | Whisper v2 | Word Error Rate (WER) | 25.7 | # 2 | |
Case Error Rate | 6.5 | # 4 | ||||
Punctuation F-1 | 50.0 | # 1 | ||||
Line break F-1 | 71.7 | # 3 | ||||
Section break F-1 | 3.1 | # 2 | ||||
Automatic Lyrics Transcription | Jam-ALT Spanish | AudioShake | Word Error Rate (WER) | 22.5 | # 1 | |
Case Error Rate | 4.1 | # 2 | ||||
Punctuation F-1 | 47.8 | # 2 | ||||
Parenthesis F-1 | 38.0 | # 1 | ||||
Line break F-1 | 82.7 | # 1 | ||||
Section break F-1 | 69.6 | # 1 | ||||
Automatic Lyrics Transcription | Jam-ALT Spanish | HTDemucs + Whisper v2 | Word Error Rate (WER) | 38.8 | # 4 | |
Case Error Rate | 7.1 | # 5 | ||||
Punctuation F-1 | 17.2 | # 5 | ||||
Line break F-1 | 56.4 | # 4 | ||||
Automatic Lyrics Transcription | Jam-ALT Spanish | HTDemucs + Whisper v3 | Word Error Rate (WER) | 61.5 | # 5 | |
Case Error Rate | 3.6 | # 1 | ||||
Punctuation F-1 | 28.7 | # 4 | ||||
Line break F-1 | 52.4 | # 5 | ||||
Automatic Lyrics Transcription | Jam-ALT Spanish | Whisper v3 | Word Error Rate (WER) | 28.6 | # 3 | |
Case Error Rate | 5.0 | # 3 | ||||
Punctuation F-1 | 41.9 | # 3 | ||||
Line break F-1 | 73.7 | # 2 |