30 April 2020; Deezer’s team published a new article around a new AI system using Spleeter, our source separation tool. We were able to build a Keyword Spotting System that uses Spleeter to detect uttered explicit keywords from singing voices. This is the first ever attempt to create an explicit content detector purely based on audio.
Tagging tracks as “explicit” can be tricky as it usually depends on cultural context and can feel quite subjective. We wanted to see if our AI tools can help change this. Our goal was to detect explicit keywords in tracks and create a tool that helps (but doesn’t replace) the person tagging explicit songs properly.
Our Director of Research, Manuel Moussallam, covers the process in more detail in a new article that was published today. The paper looks at the tool we’ve built, the challenges and whether this could play a part in the future of music and streaming.
You can check out the full article here.
For any questions or to learn more about the future of explicit lyric detection, do let me know.