I've used the electronically available Oxford Advanced Learners'
Dictionary to derive the decision tree:
                 ftp://ota.ox.ac.uk/pub/ota/public/dicts/710/    

PLEASE CONSULT THEIR LICENCE LIMITING THE USE TO NON COMMERCIAL 
APPLICATIONS.

As this dictionary provide tags, you can indicate the nature of the
word ( VERB or NVERB) to disambiguate heterophone-homographs, e.g.:

   I \VERB record a \NVERB record

Yep, this is not a full TTS and hence no part of speech tagger
inside, MBRDICO is just intended to demonstrate the automatic grapheme
to phoneme learning technique distributed in the ID3 package...

Have fun

   Vincent PAGEL
