I want to use time aligned transcription (similar to TIMIT) to train the model (instead of flat start). I am looking for alignment format which is expected by the sphinx tool. I didn't find any documentation or discussion and hence posting here. I am using sphinxtrain so have to check the internal tools which are called to train CD and CI models and transcription format expected by them. Please do point me to document or format if available any. In the mean time, I will try to get the time aligned transcription from the decoder or force aligner to get idea about the format and see if that can be used to bootstrap the model trainig.
Thanks,
rrb
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I want to use time aligned transcription (similar to TIMIT) to train the model (instead of flat start). I am looking for alignment format which is expected by the sphinx tool. I didn't find any documentation or discussion and hence posting here. I am using sphinxtrain so have to check the internal tools which are called to train CD and CI models and transcription format expected by them. Please do point me to document or format if available any. In the mean time, I will try to get the time aligned transcription from the decoder or force aligner to get idea about the format and see if that can be used to bootstrap the model trainig.
Thanks,
rrb
We support flat start only.