SPEECH RECOGNITION DATASETS: UNLOCKING THE POTENTIAL OF AUDIO AI