HyperAI超神经

首页算力平台文档资讯论文教程数据集百科 SOTA LLM 模型天梯 GPU 天梯顶会

中文

HyperAI超神经

Automatic Lyrics Transcription On Jam Alt 1

评估指标

Case-Sensitive Word Error Rate

Line break F-1

Punctuation F-1

Section break F-1

Word Error Rate (WER)

评测结果

各个模型在此基准测试上的表现结果

						Paper Title	Repository
AudioShake v3	20.9	84.3	65.3	84.8	17.3	Lyrics Transcription for Humans: A Readability-Aware Benchmark
AudioShake v1	-	80.7	59.0	77.4	22.1	Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
LyricWhiz	-	74.0	34.0	1.4	24.6	Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
LyricWhiz	28.0	74.0	34.0	1.4	24.6	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v3 +lang	41.4	72.5	41.8	2.6	36.4	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v3	42.5	71.5	41.4	2.6	37.7	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v3	-	71.5	40.9	2.6	37.7	Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
Whisper v3 +demucs	47.2	66.9	25.8	-	43.0	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v3 +demucs +lang	47.2	66.9	25.8	-	43.0	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v3 +demucs	-	66.8	23.3	-	43.0	Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
Whisper v2 +lang	43.7	65.5	34.9	11.6	39.7	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v2	-	63.0	31.3	11.2	43.8	Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
Whisper v2	47.5	63.0	31.5	11.2	43.8	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v2 +demucs	39.1	53.9	42.2	-	33.3	Lyrics Transcription for Humans: A Readability-Aware Benchmark
Whisper v2 +demucs	-	53.8	39.2	-	32.3	Jam-ALT: A Formatting-Aware Lyrics Transcription Benchmark
Whisper v2 +demucs +lang	41.3	53.4	41.8	-	35.6	Lyrics Transcription for Humans: A Readability-Aware Benchmark
OWSM v3.1 +demucs +lang	69.4	47.3	21.5	-	63.4	Lyrics Transcription for Humans: A Readability-Aware Benchmark
OWSM v3.1 +lang	74.0	42.7	22.3	-	68.6	Lyrics Transcription for Humans: A Readability-Aware Benchmark

0 of 18 row(s) selected.