These audio datasets help train and evaluate computers ability to understand, interpret and manipulate audio. Using digital audio from microphones and deep learning models, machines can accurately identify and classify sounds — and then react to what they “hear.”