Geom_density(aes(color = playlist_genre), alpha = 0.5) +įacet_wrap(~name, ncol = 3, scales = 'free') + Select(c('playlist_genre', feature_names)) %>% Overall, the songs in the dataset tend to have low acousticness, liveness, instrumentalness and speechiness, with higher danceability, energy, and loudness. $ playlist_subgenre "dance pop", "dance pop"… $ playlist_name "Pop Remix", "Pop Remix"… $ track.artist "Ed Sheeran", "Zara Lars"… You can find the code for generating the dataset in spotify_dataset.R in the full Github repo. The top four sub-genres for each were used to query Spotify for 20 playlists each, resulting in about 5000 songs for each genre, split across a varied sub-genre space. Genres were selected from Every Noise, a fascinating visualization of the Spotify genre-space maintained by a genre taxonomist. R&B, pop, and latin songs were most difficult to sort out, but R&B songs tended to be longer in duration, and latin songs were slightly more danceable than pop tracks. Low danceability helped separate out rock tracks, and high tempo provided the distinction needed to find EDM songs. Rap was one of the easier genres to classify, largely thanks to the speechiness feature. The random forest model was able to classify ~54% of songs into the correct genre, a marked improvement from random chance (1 in 6 or ~17%), while the individual decision tree shed light on which audio features were most relevant for classifying each genre: TL DR:ĭecision tree, random forest, and XGBoost models were trained on the audio feature data for 33,000+ songs. We'll look into a sample of songs from six broad genres - pop, rap, rock, latin, EDM, and R&B - to find out. With just the quantitative features, is it possible to classify songs into broad genres? And what can these audio features tell us about the qualities of each genre? Spotify has the benefit of letting humans create relationships between songs and weigh in on genre via listening and creating playlists. Those products also make use of Spotify’s vast listener data, like listening history and playlist curation, for you and users similar to you. It’s likely that Spotify uses these features to power products like Spotify Radio and custom playlists like Discover Weekly and Daily Mixes. There are 12 audio features for each track, including confidence measures like acousticness, liveness, speechiness and instrumentalness, perceptual measures like energy, loudness, danceability and valence (positiveness), and descriptors like duration, tempo, key, and mode. The Spotify Web API provides artist, album, and track data, as well as audio features and analysis, all easily accessible via the R package spotifyr. Is it possible to classify songs into broad genres? And what can quantitative audio features tell us about the qualities of each genre? Exploring Spotify's audio features “folk rock,” but rather the listener knows it when they hear it. Musical genre is far from black and white - there are no hard and fast rules for classifying a given track or artist as “hard rock” vs.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |