Playlist generation is an important task in music information retrieval. While previous work has treated a playlist
collection as an undifferentiated whole, we propose to build
playlist models which are tuned to specific categories or
dialects of playlists. Toward this end, we develop a general
class of flexible and scalable playlist models based upon
hypergraph random walks. To evaluate the proposed models, we present a large corpus of categorically annotated,
user-generated playlists. Experimental results indicate that
category-specific models can provide substantial improvements in accuracy over global playlist models.
1. Hypergraph models of
playlist dialects
Brian McFee Lab
Center for Jazz Studies/LabROSA
Columbia University ROSA
Laboratory for the Recognition and
Organization of Speech and Audio
Gert Lanckriet
Electrical & Computer Engineering
University of California, San Diego
7. Data collection
http://www.artofthemix.org/
Started in 1998, users upload and share playlists
[Ellis, Whitman, Berenzweig, and Lawrence, ISMIR 2002]
8. The data: AotM-2011
• 98K songs indexed to Million Song Dataset
• 87K playlists (1998-2011), ~210K contiguous segments
• 40 playlist categories, user meta-data available
9. # Playlists per category
Mixed genre
Theme
Rock-pop
Alternating DJ
Indie
Single artist
Romantic
Road trip
Depression
Punk
Break-up
Narrative
Hip-hop
Sleep
Dance-house
Electronic
Rhythm & blues
Country
Cover
Hardcore
Rock
Jazz
Folk
Ambient
Blues
100 1000 104 105
10. # Playlists per category
Mixed genre
Theme
Rock-pop
Alternating DJ
Indie
Single artist
Romantic
Road trip
Depression
Punk
Break-up
Narrative
Hip-hop
Sleep
Dance-house
Electronic • Majority of playlists are Mixed genre
Rhythm & blues
Country
Cover
Hardcore • Remaining categories:
Rock
Jazz
Folk
contextual/mood, genre, other
Ambient
Blues
100 1000 104 105
11. Our goals
• Which categories can we model? Are some harder than others?
• Which features are useful for playlist generation?
• Do transitions matter? Are some categories less diverse?
21. Connecting the dots...
• Random walk on a hypergraph
- Vertices = songs
- Edges = subsets
• Learning: optimize edge weights from example playlists
22. Connecting the dots...
• Random walk on a hypergraph
- Vertices = songs
- Edges = subsets
• Learning: optimize edge weights from example playlists
• Sampling is efficient, edge labels provide transparency
27. Evaluation protocol
• Repeat x10:
- Split playlist collection into 75% train/25% test
- Learn edge weights on training playlists
- Evaluate average likelihood of test playlists
• Compare gain in likelihood over uniform shuffle baseline
28. Experiment 1: global vs. categorical
• Fit one model per category
• Fit one global model to all categories
• Test on each category and compare likelihoods
• Question:
When does categorical training improve accuracy?
29. Experiment 1: global vs. categorical
Unifo
rm
ALL
Mixed Global model
Theme Category-specific
Rock-pop
Alternating DJ
Indie
Single artist
Romantic
Road trip
Punk
Depression
Break up
Narrative
Hip-hop
Sleep
Electronic
Dance-house
R&B
Country
Cover songs
Hardcore
Rock
Jazz
Folk
Reggae
Blues
0% 5% 10% 1 5% 20% 25%
Log-likelihood gain over uniform shuffle
30. Experiment 1: global vs. categorical
Unifo • Largest gains for genre playlists
rm
ALL
Mixed • No change for "hard" categories
Global model
Theme Category-specific
Rock-pop
Alternating DJ
(e.g., Mixed, Alternating DJ, Theme)
Indie
Single artist
Romantic
Road trip
Punk
Depression
Break up
Narrative
Hip-hop
Sleep
Electronic
Dance-house
R&B
Country
Cover songs
Hardcore
Rock
Jazz
Folk
Reggae
Blues
0% 5% 10% 1 5% 20% 25%
Log-likelihood gain over uniform shuffle
31. Experiment 1: learned edge weights
ALL
Mixed
Theme
Rock-pop
Alternating DJ
Indie
Single Artist
Romantic
RoadTrip
Punk
Depression
Break Up
Narrative
Hip-hop
Sleep
Electronic music
Dance-house
Rhythm and Blues
Country
Cover
Hardcore
Rock
Jazz
Folk
Reggae
Blues
Audio CF Era Familiarity Lyrics Tags Uniform
32. Experiment 2: continuity?
• Do we need to model playlist continuity?
edge weights
songs
• Simplified model:
- ignore transitions
- choose each edge IID
exp. prior
playlists
• Question:
Are some categories more diverse than others?
33. Experiment 2: continuity
Unifo
rm
ALL
Mixed Global model
Theme Category-specific
Rock-pop
Alternating DJ
Indie
Single artist
Romantic
Road trip
Punk
Depression
Break up
Narrative
Hip-hop
Sleep
Electronic
Dance-house
R&B
Country
Cover songs
Hardcore
Rock
Jazz
Folk
Reggae
Blues
-15% -10% -5% 0% 5% 10% 15% 20%
Log-likelihood gain over uniform shuffle
34. Experiment 2: continuity
Unifo
rm
ALL
Mixed Global model
Theme Category-specific
• Most categories exhibit both
Rock-pop
Alternating DJ
Indie
continuity AND diversity
Single artist
Romantic
• Transitions are important!
Road trip
Punk
Depression
Break up
Narrative
Hip-hop
Sleep
Electronic
Dance-house
R&B
Country
Cover songs
Hardcore
Rock
Jazz
Folk
Reggae
Blues
-15% -10% -5% 0% 5% 10% 15% 20%
Log-likelihood gain over uniform shuffle
35. Example playlists
Rhythm & Blues
EDGE SONG
70s & soul Lyn Collins - Think
Audio #14 & funk Isaac Hayes - No Name Bar
DECADE 1965 & soul Michael Jackson - My Girl
Electronic music
EDGE SONG
Audio #11 & downtempo Everything but the Girl - Blame
DECADE 1990 & trip-hop Massive Attack - Spying Glass
Audio #11 & electronica Björk - Hunter
36. Conclusions
• Category-specific models outperform global playlist models.
• Continuity matters!
• Proposed model is simple, efficient, and transparent
• AotM-2011 dataset available now!
http://cosmal.ucsd.edu/cal/projects/aotm2011