4 ms·
This only thing unusual about this problem for classical music is the extremely large repertoire of public domain works. The article vaguely implies that there
by nullc 6y ago
This only thing unusual about this problem for classical music is the extremely large repertoire of public domain works.
The article vaguely implies that there is also an issue of miscalibration of the matching due to emphasising commercial pop-music but I've never seen any evidence of that and wouldn't expect it from how these systems work. In particular, I've found various contentid services to be effective at identifying classical music-- so clearly their database contain plenty of it. I expect that this claim is pure speculation on the part of the article's author.
- msandford 6y agoI think the idea of miscalibration of matching isn't that the system can't ID classical, it's that the system doesn't know that classical should have different rules apply. A lot of pop music "sounds the same" but the songs are actually different in meaningful ways note-wise. When you're dealing with classical music it's all the same notes! Just different people playing it. So probably what's happening (some-how, some-way) is that the ID mechanism is trained to find songs that are "the same" and in pop music they're only "the same" if they're the _same song_. But in classical there are many songs that are different that are the _same song_ and so the model which was trained with one set of assumptions completely fails when a different set of assumptions are in play. And because classical doesn't register in a big way to the people in charge this doesn't get fixed because it only takes the algorithm from 99.85% accurate to 99.72% accurate which as we all know, is meaningless!
- zozbot234 6y ago> So probably what's happening (some-how, some-way) is that the ID mechanism is trained to find songs that are "the same" and in pop music they're only "the same" if they're the _same song_. Right. Pop music has a lot of variation in timbre even between 'covers' of the same composition, so if the notes match and the overall timbre matches it's a safe bet that you're looking at two instances of the exact same recording, or at most two versions by the same performer. In classical that just isn't true, the timbre of acoustic instruments is basically fixed and the notes are defined by the original composition. There's just too little variation for an auto-id system to work with.