Nice work ! So The OP said here that this is based on a “semantic” enhancing of an (already rich) 35k tracks DB. Music platforms in general do not allow peopl

Nice work ! So The OP said here that this is based on a “semantic” enhancing of an (already rich) 35k tracks DB. Music platforms in general do not allow people to describe/annotate tracks verbally (people would love to "let it out" !). The day they will decide to implement an NLP search similar to this, for up to 100 million tracks [1], they will regret it. Album reviews is the only "semantically rich" AND "widely used" music description I can think of. Last.fm has some interesting data on the finer track level (wiki, comments, tags), but it could have been way, way richer [1] https://www.apple.com/newsroom/2022/10/celebrating-100-million-songs/ https://www.apple.com/newsroom/2022/10/celebrating-100-milli...