For pocketsphinx, I'd like to create a .dic file from a text corpus with
phones for word pronunciation, including best guesses for unknown words. The
online LMTool works very well for my purposes but I am looking for an option
that is self-contained on the device.
So far I've found logios (but I can't use perl) and the LetterToSoundImpl.java
class of FreeTTS (but I can't use Java) and I've noticed that flite can also
make a best guess in -ps mode, but unlike the online LMTool it offers a single
possible pronunciation, whereas the online LMTool frequently comes up with
alternative pronunciations which are pretty good. Is there a c-based (or
c++-based) tool for deriving pronunciations of unknown words with as many
variations in possible pronunciation as the online LMTool? I can do a certain
amount of perl->c porting but what was in the dictionary-looking-up part of
logios struck me as being more than I could take on right now. Thanks very
much for any pointers.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello,
For pocketsphinx, I'd like to create a .dic file from a text corpus with
phones for word pronunciation, including best guesses for unknown words. The
online LMTool works very well for my purposes but I am looking for an option
that is self-contained on the device.
So far I've found logios (but I can't use perl) and the LetterToSoundImpl.java
class of FreeTTS (but I can't use Java) and I've noticed that flite can also
make a best guess in -ps mode, but unlike the online LMTool it offers a single
possible pronunciation, whereas the online LMTool frequently comes up with
alternative pronunciations which are pretty good. Is there a c-based (or
c++-based) tool for deriving pronunciations of unknown words with as many
variations in possible pronunciation as the online LMTool? I can do a certain
amount of perl->c porting but what was in the dictionary-looking-up part of
logios struck me as being more than I could take on right now. Thanks very
much for any pointers.
It would be nice to apply fsttools to do letter to sound convertions:
http://www.clsp.jhu.edu/wiki2/Pronunciations_Lab
it's should be trivial to get n-best then.
That would be cool, I'm checking it out - thank you. What tool does the non-
logios version of the online tool use to generate pronunciations?
Whoops, silly me, they are matches out of the dictionary.