search

IBM / Train-Custom-Speech-Model

Create a custom Watson Speech to Text model using specialized domain data
https://developer.ibm.com/patterns/customize-and-continuously-train-your-own-watson-speech-service/
Apache License 2.0
59 stars 42 forks source link

Deleting and adding 2nd corporas #63

Open Twanawebtech opened 5 years ago

Twanawebtech commented 5 years ago

While playing with the app, I noted a few things.

1 I added a new corpora language model than did the training and all worked fine. 2 Next then I added second corpora using the same sound file and updated a few other things. Then when I do training again and use the base model, I only see my corpora1 taken effect. 3_ Next then I deleted all corporate yet still able to run the custom language model and getting corpora1 output. There seems to be a bug with adding 2nd corpora's and removing corpora's.

4_ Lastly, when all corpora's deleted then I see the pendding status for my model training.

image

tonanhngo commented 5 years ago

@Twanawebtech (2) should work, we will try to replicate the error you are seeing. (3) I think after deleting the corpora, you still have the custom model that has been trained, so it should continue to be available. You can reset the custom model to get back to the starting point and this will wipe out the trained model. The GUI doesn't support this feature, but you can run the command line "reset_language_model.py". (4) This is normal: the meaning of the "pending" state is that it is waiting for you to add corpora so that training can be kicked off.