building the mlf from the front end

[Pilar O]

Hi Simon,

I’m plugging in my set of LTS rules to the script “build_unitsel.scm” to get an mlf to start the process of building the utts for the voice.
I wanted to confirm something with you:
when we build the mlf are we actually ignoring the punctuation symbols?

My rules are acting weird with punctuation symbols because I think this set of LTS does not have a line that the others have, I think the other dicts only take strings made up with letters (they have regex near the start “(string-matches dcword “[a-z\-]*”)”).

My current front end can’t do that because it need the “‘” symbols to get the orthographic stress. But I don’t find a way of make it use the punctuation/phrase break tree when is building the mlf in the function “make_initial_phone_labels” (and the used function “setup_phoneset_and_lexicon” where it actually calls the lexicon and the LTS to build the phone string) in the mentioned script…

Therefore… should I just remove the punctuation symbols from my utts.data script when I build the mlf file? (seems the easiest thing to do, and apparently it would not have any effect on the mlf).

Thank you!

[Simon]

Stripping problematic punctuation from utts.data should be OK in your first build of this system. Come back and solve this problem later.

[Author]

Posts