Quantification of stylistic differences in human- and ASR-produced transcripts of African American English