Robert Brown blogged a couple weeks ago about New utilities for Speech Server development. i finally got around to playing with them. the 1st mistake i made was trying to run them on my dev machine, when i needed to run them on my Speech Server machine. the rest of this post will just be a quick run through of what the utilities do. they are all command line
just takes a file list of words and returns a file list of pronunciations for those words. very similar to CMUDict, but with the addition of the 'ax' phoneme. it can return multiple prons for the same word (marked with an iter). it has a lexicon of prons and can create a pron for a word not in the lexicon (marked with *). if it cannot gen a pronunciation then it spits out the pron for 'bogus pronunciation'. wish that this would return which phonemes should be stressed. and i hope we get programmatic access to this functionality as well.
dumps out some stats for grammar files in .grxml and .cfg format. i ran this against a couple different grammar files and looked at the results, but dont see how this is useful to me? maybe if i was editing the grammars by hand, instead of using the IDE. and it lets you easily see the performance diff between .grxml and .cfg files.
this takes a grammar file and a wav file as input (optionally a rule name). it then does the speech recognition of the wav to the grammar and outputs the SML result. this tool could be used to create unit tests for your grammars.
this returns the human readable error message for a SAPI error code. as a developer, i just want this functionality in my IDE. the doc says its intended purpose is to be run against error codes in the Speech Server logs.