Speech Recognition on Long-form video (test)
2.08Avg WERSeed-ASR
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Seed-ASRSFT Protocol=long-form SFT2024.07 | 2.08 | 1.44 | 1.96 | 1.95 | 2.56 | 2.31 | |
| Seed-ASRSFT Protocol=short-form SFT2024.07 | 2.28 | 1.48 | 1.99 | 2.31 | 2.64 | 2.73 | |
| Transducer-based E2E Model2024.07 | 3.92 | 2.83 | 3.8 | 3.8 | 4.22 | 4.66 | |
| Paraformer-large2024.07 | 5.97 | 5.78 | 5.36 | 5.8 | 6.87 | 5.96 |