But the tech is pretty f%#&ing far from perfect.
Deezer says it has developed a system to automatically detect explicit content in songs, and will soon publish a research paper about its findings.
The streaming service’s head of research and development, Manuel Moussallam, detailed the process of developing this system in a lengthy Medium post. Currently, he notes that only humans at record labels and streaming services are in charge of deciding whether a song should be tagged as explicit or not — and a streamlined system for doing so doesn’t exist yet.
Labels sometimes deliver songs to Deezer with the “explicit” tag, but more often than not, Moussallam says, Deezer is tasked with figuring out whether the song contains explicit content. The process “requires a high-level understanding of cultural expectations and involves a lot of subjectivity,” he says.
Moussallam adds that any explicit content-detecting service must be tested on all genres equally, to help remove a genre bias: “If you’re not careful when designing your experiment,” he says, “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.”
Moussallam says his research team was able to build a modular system that decides whether a song is “explicit” based on lyrical keywords. But, he warns, “we do NOT consider it fit for tagging songs as explicit in a fully automated manner.”
The full research paper will be published at the upcoming International Conference on Acoustics, Speech, and Signal Processing, which will be held virtually, May 4-8. For now, you can read Moussallam’s Medium post here.