Longmans Pronunciation Dictionary (LPD) is expensively available in that
form (ask John Wells), and one of the Oxford dictionaries (ask Michael
Ashby). I'm not sure whether there is now anything in the public domain,
freely available for speech technologists to use. A request to the SALT
list might be fruitful, but you would need to be more precise about what you
need.
Decisions about syllabification and stress marking, as I'm sure you are
aware, vary from transcriber to transcriber, depending on their theoretical
position. For example, LPD marks secondary stresses which occur before the
primary stress, but any rhythmic prominence (John calls it tertiary
stress) which occurs AFTER the primary stress is left unmarked. Some
dictionaries mark these too, and in practice, for synthesis we will need to
identify strong vs weak syllables which occur in this position. Syllable
boundary-placement, as I mentioned earlier, is theory-dependent, and any
structure implied by a dictionary's transcription will almost certainly
have to be modified for our purposes...
That said, for your immediate parser development needs it wouldn't matter
too much how syllables were marked, as long as you could count them and
differentiate strong from weak etc.
Jill