Long-form Transcription on Earnings-22
10.88WERTRADE
Evaluation Results
| Method | Links | |
|---|---|---|
| TRADEdecoding=streaming, chunk size=5,120 ms2026.06 | 10.88 | |
| Whisper large-v22022.12 | 12.6 | |
| Whisper medium2022.12 | 13.2 | |
| Whisper medium.en2022.12 | 13.3 | |
| Whisper large2022.12 | 13.4 | |
| FC-XXLDecoder=CTC, Train Dataset=ASR Set ++2023.05 | 13.69 | |
| Canary-1B-v2decoding=parallel chunks2026.06 | 13.93 | |
| Whisper small.en2022.12 | 14 | |
| FC-XXLDecoder=RNNT, Train Dataset=ASR Set ++2023.05 | 14.11 | |
| FC-XLDecoder=CTC, Train Dataset=ASR Set ++2023.05 | 14.14 | |
| Whisper small2022.12 | 14.3 | |
| FC-XLDecoder=RNNT, Train Dataset=ASR Set ++2023.05 | 14.78 | |
| FC-XLDecoder=RNNT, Train Dataset=ASR Set2023.05 | 16.37 | |
| FC-XXLDecoder=CTC, Train Dataset=ASR Set2023.05 | 16.44 | |
| Whisper base.en2022.12 | 16.6 | |
| FC-XXLDecoder=RNNT, Train Dataset=ASR Set2023.05 | 16.66 | |
| stt_en_conformer_ctc_large2022.12 | 17.6 | |
| FC-XLDecoder=CTC, Train Dataset=ASR Set2023.05 | 17.89 | |
| Whisper base2022.12 | 18.4 | |
| Fast Conformervariant=FT+LCA+GT2026.06 | 19.49 | |
| stt_en_conformer_transducer_xlarge2022.12 | 19.7 | |
| Whisper tiny.en2022.12 | 22 | |
| Whisper tiny2022.12 | 24.4 | |
| wav2vec2-large-960h-lv60-self2022.12 | 28 | |
| hubert-large-ls960-ft2022.12 | 30.6 | |
| wav2vec2-large-robust-ft-libri-960h2022.12 | 31 | |
| hubert-xlarge-ls960-ft2022.12 | 31.3 | |
| wav2vec2-large-960h2022.12 | 40.1 | |
| wav2vec2-base-960h2022.12 | 46.8 | |
| wav2vec2-base-100h2022.12 | 57.1 |