Output textGrid needs Utterance Tier

[dinemamadou]

Hi @mmcauliffe, Is there a flag or configuration I can use to tell the MFA to not remove the utterance tier during the alignment process or to have the MFA work on a designated textGrid tier or maybe create a copy of the existing tier and work on that one instead?

I was able to train acoustic models for Baatonum, Dan and Ede Chaabe, but the output textGrids lack the utterance tier, which is very important for the project I'm working on. The project is about declarative questions, where the segmental and word makeup of statements and questions are the same. As such, I need the utterance labels to distinguish between sentence type.

Any help is welcome.

Thanks in advance, Deen

[mmcauliffe]

Sure, I can add some functionality for that in the next release. I'll have to think about how to structure it and expose it, but can probably at least have options for exporting the utterance's sanitized text and the normalized text. It sounds like you'd probably need the un-sanitized text though, right? I can probably refactor so that the original text gets saved and is then exportable too.

[tjmahr]

Duplicate of #414, so someone should close.