1. Tony Robinson's BEEP dictionary: svr-ftp.eng.cam.ac.uk/comp.speech
2. The Longman Pronounciation Dictionary
I've had a look at the BEEP dictionary and found out that it is based on
OALD (Oxford Advanced Learners' Dictionary) and MRC (Medical Research
Council) compiled at Imperial. It has 250,000 word forms but only a portion
of which are annotated for stresses. I don't know how large the portion is
but presume it's the OALD component, that is, circa 60,000 base forms.
Ideally, we'd need somthing more comprehensive than this. As a second issue,
the syllable divisions are marked only in the IPA representations, not in
the actual word form. It's not clear yet to what extent the IPA
representations can be mapped unto the lexical forms.
According to Jill, the Longman dictionary is expensive. I suppose it only
has lemmas rather than all the inflected forms. It's also fairly small (in a
computer sense), about 75,000 entries.
More suggestions are welcome and I'll post more as I move along.
Alex