I checked the WSJ generic model and see there are some "garbage" model, like laugh, noise model there. Now I have the grammar for the ASR task, and I expect there will be some noise and laugh in the speech, can I indicate them in the grammar?
Thank you.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Fillers are automatically inserted in grammar after each word. You can tune filler probability with -fillprob configuration option. Other relevant option is -fsgusefiller (enabled by default).
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I checked the WSJ generic model and see there are some "garbage" model, like laugh, noise model there. Now I have the grammar for the ASR task, and I expect there will be some noise and laugh in the speech, can I indicate them in the grammar?
Thank you.
Fillers are automatically inserted in grammar after each word. You can tune filler probability with -fillprob configuration option. Other relevant option is -fsgusefiller (enabled by default).