search

alphacep / vosk-api

Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Apache License 2.0
7.92k stars 1.1k forks source link

Vosk-api python Vietnamese model error rate #879

Closed rbsongcan closed 2 years ago

rbsongcan commented 2 years ago

Hello, thanks for this great Vosk ASR engine. I'm using the Vietnamese model and want to learn how to lower the error rate. Can someone tell me where to start ? And which project of the current Vietnamese model please?

nshmyrev commented 2 years ago

Can someone tell me where to start?

Start with collecting diverse transcribed speech data from various sources. Youtube is a good candidate. You need at least 1000 hours.

And which project of the current Vietnamese model please?

We don't have a specific project for the languages, we just released training recipe though: https://github.com/alphacep/vosk-api/tree/master/training

rbsongcan commented 2 years ago

@nshmyrev thank you for the answer. I will try to learn the training setup. Hope there will be more document updated in README.

rbsongcan commented 2 years ago

Provide this info for who has the same concern. What I'm looking for is in this link: https://alphacephei.com/vosk/models