With so much new music created and uploaded to streaming platforms daily, those platforms face the potential problem of identifying songs that include explicit content or lyrics. Those who wish to filter explicit content need it to be identified within the platform in order to avoid it. Detecting explicit content amid the vast expanse of music can prove to be tricky for music platforms, and using human ears is a process that would become a virtually endless task.
Deezer has been working on a solution. Its new tool, Spleeter, uses machine learning in order to separate and identify explicit lyrics within songs, allowing the platform to tag items with adult content more easily. The system extracts vocals and matches dictionary entries of harsh language before submitting them for human review, saving time for humans who would instead be listening through entire songs to find anything that could be deemed explicit.
Spleeter's learning can be problematic at times, and Deezer will have to be cautious in ensuring that the system doesn't develop any biases. “If you’re not careful when designing your experiment, you may end up with a system that instead of detecting explicit lyrics, will detect rap songs, just because they have on average, a higher probability of containing explicit lyrics,” Deezer says.
The platform is hoping to get Spleeter to a point where it can be used completely hands-free without any human intervention or machine-learned bias.
H/T: Digital Music News