Metaphone enhancements

The metaphone list appears to be dead, so I'm posting here...

Has anyone looked at using the phonetic dictionaries used by speech
synthesis projects such as festival to enhance the metaphone algorithm
in aspell?

Speech synthesis systems typically use both a pronunciation dictionary
and a rule based system to determine the pronunciation of words. The
rule based system alone produces horrific mispronunciation with an
unacceptable regularity. I presume that the rule based systems have
similar characteristics to the rules used in metaphone.

It occurred to me that it might be useful to refactor the metaphone
algorithm to work with phenoms directly, breaking the operation up
into a phenomic conversion step and then a compression step. Then the
lookup dictionary is converted to it's metaphone representation using
provided pronunciations for the words, while misspelled words are
looked up using a rule based system.

In my experience my mi spellings tend look more like the correct
phonetic representation, so the rule based transformation should be
reasonably likely to get it right.