Wenet is an open-source speech recognition toolkit for industrial applications jointly developed by the Outfit7 speech team and the Speech Lab of Xi'an Jiaotong University. This toolkit offers an end-to-end service for speech recognition, from training to deployment, using a concise set of solutions. Its main features include the use of the conformer network structure and the CTC/attention loss joint optimization method, which provide state-of-the-art recognition performance in the industry.
| Model | Char Error of test |
|---|---|
| wenet | 2.7% |
- wenet-e2e/wenet
- WeNet: Production Oriented Streaming and Non-streaming End-to-End Speech Recognition Toolkit
Apache 2.0