Another question for you, Halle.
It would be very useful to be able to write automatic tests that verify any one particular language model recognises pre-recorded segments of speech. However, the only function that is able to perform this task,
runRecognitionOnWavFileAtPath:, DOES NOT RETURN A HYPOTHESIS!
In order for a unit test to verify that a wav file is correctly interpreted as a the expected sentence, the only way I see it being possible is to capture NSLog into a file first (see http://parmanoir.com/Redirecting_NSLog_to_a_file) and then opening that file to make assertions afterwards.
The problem with that approach is that while I can stomach the horribleness, XCode relies on NSLog to be written to the console to provide accurate test reporting so even redirecting NSLog is not an attractive option.
Please return a hypothesis or something so that automated testing can be done. Thanks!