search

dopefishh / praatalign

A Praat plug-in for performing interactive phonetic forced alignment
MIT License
26 stars 5 forks source link

Numbers and symbols in dict #9

Open ERSXS opened 7 years ago

ERSXS commented 7 years ago

It seems that the script doesn't know what to do when dictionary entries contain numbers or IPA symbols, although some punctuation is fine. When using Align current interval, I get this error message:

HVite failed with the following error: ERROR [+8251] ReadLattice: Word 3: not in dict ERROR [+3210] DoAlignment: ReadLattice failed FATAL ERROR - Terminating program C:\htk\bin.win32\HVite.exe

ERSXS commented 7 years ago

So, nevermind about the IPA, but it seems the issue is a mismatch between what your cmu2praatalign.py script produces and what your MONOPHONES file lists: in cmu2praatalign, you translate 'ZH' to '3' instead of 'Z' and 'ER' to '3:' — I think '3:' corresponds to maybe 'P3:' in MONOPHONES?

dopefishh commented 7 years ago

Thanks, this is probably a mistake indeed. If you want you can patch this in the English parameter directory. Otherwise I will patch it asap.

ERSXS commented 7 years ago

What need to be done to patch it? When I try replacing instances of 'P3:' with '3:' in DICT, MONOPHONES, and MMF, Praat gives me the error

HVite failed with the following error: ERROR [+8050] ReadDict: Probability malformed 3: ERROR [+8013] ReadDict: Dict format error ERROR [+3213] Main: ReadDict failed FATAL ERROR - Terminating program C:\htk\bin.win32\HVite.exe

dopefishh commented 7 years ago

can you send me a small example that uses this configuration?

ERSXS commented 7 years ago

Sure! par.english.zip

Changes were made to DICT, MMF, and MONOPHONES