Long-form Transcription on Earnings-21
9.7WERWhisper large-v2
Evaluation results
| Method | Links | |
|---|---|---|
| Whisper large-v22022.12 | 9.7 | |
| TRADEdecoding=streaming, chunk size=5,120 ms2026.06 | 9.75 | |
| Whisper medium.en2022.12 | 10.2 | |
| Whisper medium2022.12 | 10.3 | |
| Whisper large2022.12 | 10.3 | |
| Whisper small.en2022.12 | 10.8 | |
| Whisper small2022.12 | 11.1 | |
| Whisper base.en2022.12 | 12.5 | |
| stt_en_conformer_ctc_large2022.12 | 12.6 | |
| Whisper base2022.12 | 13.5 | |
| Fast Conformervariant=FT+LCA+GT2026.06 | 13.84 | |
| stt_en_conformer_transducer_xlarge2022.12 | 16.2 | |
| Whisper tiny.en2022.12 | 17 | |
| Whisper tiny2022.12 | 18.7 | |
| wav2vec2-large-960h-lv60-self2022.12 | 21.7 | |
| wav2vec2-large-robust-ft-libri-960h2022.12 | 23 | |
| hubert-large-ls960-ft2022.12 | 23 | |
| hubert-xlarge-ls960-ft2022.12 | 23.1 | |
| wav2vec2-large-960h2022.12 | 30.4 | |
| ChunkFormercontext size [latt, c, r]=[128, 64, 128]2025.02 | 31.73 | |
| ChunkFormercontext size [latt, c, r]=[128, 256, 128]2025.02 | 32.04 | |
| ChunkFormercontext size [latt, c, r]=[256, 128, 128]2025.02 | 32.05 | |
| Efficient Conformerdecoding_mode=full-context2025.02 | 36.64 | |
| wav2vec2-base-960h2022.12 | 37.3 | |
| Squeezeformerdecoding_mode=full-context2025.02 | 38.09 | |
| ChunkFormerdecoding_mode=full-context2025.02 | 38.67 | |
| Conformerdecoding_mode=full-context2025.02 | 39.39 | |
| wav2vec2-base-100h2022.12 | 45.7 |