A project looking at the error rates of VOSK ASR on African American English using the CORAAL dataset
# AAVE_VOSK
A project looking at the error rates of VOSK ASR on African American English using the CORAAL dataset
[THIS REPO IS STILL IN PROGRESS]
We know that non-standard dialects of english are transcribed less accurately. This repo is a test of VOSK's large english model on a dialect of AAE spoken in Atlanta.
Further analysis incoming looking at which types of errors vosk makes (ie. re-adding copulas, they -> their...)