Investigating the Effect of Sound-Event Loudness on Crowdsourced Audio Annotations

How well do people label soft sounds in audio data? Does it depend on factors other than loudness, such as the type of sound? Does it matter for training/evaluating machine learning models?

NYU CUSP Senior Research Scientist Justin Salamon examines these questions in new research on “Investigating the Effect of Sound-Event Loudness on Crowdsourced Audio Annotations.”