I'm trying to do adaptive student modeling in a language
learning context, and there are too many biphones for
quick adaptation, at least in the raw TIMIT corpus. I want
to map down to the fewer phonemes of the ARPABET or the IPA;
hav you or anyone that you know of been working on that?
I will summarize replies made directly to me to both of
these mailing lists to help cut down on traffic.
Sincerely,
:James Salsman
P.S. There is more information on TIMIT at
http://www.ldc.upenn.edu/ldc/catalog/html/speech_html/1timit.html
and on the phonetic codes at
http://www.ldc.upenn.edu/doc/timit/phoncode.doc