Closed rolandomunoz closed 4 years ago
No, there's no easy way to exploit word-aligned data. First, the model works at the phones level anyways so that's not enough information even to initialize the model informatively; secondly, there's no easy way to feed it phone information.
Just train it at the utterance level instead, using the -t
flag. It
almost always works.
On Sat, Feb 15, 2020 at 8:06 AM Rolando Muñoz notifications@github.com wrote:
I have a few thousand words aligned in TextGrid files. Is there a way to use these files as input for segment alignment?
Thank you!
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/prosodylab/Prosodylab-Aligner/issues/80?email_source=notifications&email_token=AABG4OPDUJRMCIJSGHYCKOTRC7SE7A5CNFSM4KVYRJW2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4INYTBJQ, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABG4OLS437WWRQRMNBRYFLRC7SE7ANCNFSM4KVYRJWQ .
I have a few thousand words aligned in TextGrid files. Is there a way to use these files as input for segment alignment? I mean, using the time boundaries as the start and end for the first and last segment, instead of creating TextGrid files from the scratch.
Thank you!