Closed dvdsk closed 4 years ago
not doable, current approach uses google knowledge graph + spacy (python lib) natural language parsing to filter out works of art. The idea was to then parse the wiki page of that work of art to figure out what its setting is (medieval, modern, detective, magic etc). However the first stage already gives to much false positives with unknown tracks.
This might be solved by only running this algorithm if a given track can be recognized based on its musical fingerprint.
search Wikipedia and or other text sources for keywords to further categorize music themes (think medieval, modern, detective, atmospheric etc)