Investigating the Corpus Phonetics Pipeline Applied to Diverse Speech Data