Taste or Addiction?: Using Play Logs to Infer Song Selection Motivation
1. Taste or Addiction?:
Using Play Logs to Infer Song Selection Motivation
Kosetsu Tsukuda and Masataka Goto
National Institute of Advanced Industrial Science and Technology (AIST), Japan
PAKDD 2017
2017/05/26
2. 2Play logs of online music services
Online music service
Play logs of Last.fm
User ID Artist Song title Timestamp
55 Madonna Crazy For You 1/5/2017 11:24:35
141 The Beatles Let It Be 1/5/2017 11:28:16
55 Lady Gaga Poker Face 1/5/2017 11:28:30
โฎ
3. 3Userโs song selection motivation
๐ก๐ก
๏ฌ By extracting a userโs logs, his play history can be generated:
it is observable when he listens to music and what he listens to
๏ฌ It is not observable why the user played each song
I played this song because โฏ
Research goal
Given a userโs play history,
infer the userโs song selection motivation of each played song
4. 4Factor 1: userโs usual taste in music
๐ก๐ก
๏ฌ User has his/her own taste in music
๏ฌ Hypothesize that such taste affects useโs song selection
I played these songs because I like rock music.
rockrockrock
5. 5Factor 2: userโs addiction to an artist
๐ก๐ก
๏ฌ User can be addicted to an artist regardless of his taste in music
(he watched the artist on TV or his friend recommended the artist)
๏ฌ Hypothesize that such addiction affects useโs song selection
My friend recommended an album of Lady Gaga.
I played these songs because I became addicted to her.
L.G. L.G.
6. 6What can our model do?
๐ก๐ก
Taste Addiction Taste
Taste: 73%
Addiction:
27%
Taste/Addiction ratio Taste
Rock: 45%
Hip-hop:
19%
Addiction
Lady Gaga:
45%Madonna:
32%
Our model can estimate userโs song selection motivation and
userโs statistics (e.g., taste in music and addicted artists)
7. 7What can our model do?
Our model can reveal usersโ addiction-based music listening behavior
๏ฌ Users have high addiction degree in the morning
๏ฌ Users have low addiction degree on the weekend
๐ก๐ก
Taste Addiction Taste
๐ก๐ก
Addiction Taste
๐ก๐ก
Taste Addiction Addiction
โฎ
9. 9Overview
๏ฌ Estimate userโs song selection motivation by using a topic model
๏ฌ Our model builds on the one proposed by Zheleve et al.
in WWWโ10 that considers only userโs taste in music
Model by Zheleve et al.
Taste
Our model
Taste Addiction+
10. 10Session
๐ก๐ก
๏ฌ Divide userโs play logs into sessions
๏ฌ If the time gap between two songs is less than 30 min,
they belong to the same session
41 min 135 min
Session 1 Session 2 Session 3
๐ก๐ก
๐ก๐ก
11. 11Song generative process by Zheleve et al.
๐ก๐ก
๏ฌ User has a topic distribution that represents the userโs taste
๏ฌ One topic is assigned to each session based on the userโs taste
๏ฌ Artist is selected based on the artist distribution of the topic
Topic distribution = Userโs taste
Prob
Topic 1 2 3 4 5
โฏ
Topic 5
Artist distribution of topic 5
Prob
โฏ
Eminem
JayZ
J.Cole
ICECUBE
2Pac
Eminem
J.Cole
Eminem
Topic 5 Topic 2
JayZ
Annekei
Annekei
12. 12Graphical model
Topic distribution = Userโs taste
Prob
Topic 1 2 3 4 5
โฏ
Artist distribution of topic 5
Prob
โฏ
Eminem
JayZ
J.Cole
ICECUBE
2Pac
13. 13Our Model
๏ฌ User has a taste-addiction weight that affects his song selection
๏ฌ In an addiction mode, the user directly chooses a song
๏ฌ User has an artist distribution that represents the userโs addiction
Prob
Taste
Addiction
Taste-addiction weightAddicted artist distribution
Prob
โฏ
LadyGaga
DianaRoss
BrunoMars
Adele
LMFAO
14. 14Song generative process in our model
For each session
๏ฌ Choose a topic
For each song in the session
๏ฌ Choose a mode (taste or addiction)
๏ฌ If taste mode, choose an artist according to the topic
๏ฌ Else if addiction mode, choose an artist
from addiction artist distribution
๐ก๐ก
Topic 5
Eminem
LadyGaga
LadyGaga
Taste
Addiction
Addiction
Addiction
Taste
Addicted artist distribution
Prob
โฏ
LadyGaga
DianaRoss
BrunoMars
Adele
LMFAO
Adele
J.Cole
17. 17Research Question
Is adopting two factors, which are usersโ taste in music and
addiction to artists, effective to model music listening behavior?
VS
Model by Zheleve et al.
Taste
Our model
Taste Addiction+
18. 18Dataset 1
๏ฌ Play logs of a music download service in Japan
๏ฌ Consists of 10 weeks logs from 1/1/2016 to 10/3/2016
๏ฌ 8 weeks are used as training data and 2 weeks as test data
Users Artists Logs for train Sessions for train Logs for test Sessions for test
13,986 6,431 331,437 82,427 57,126 13,516
โฎ
๐ก๐ก
๐ก๐ก
๐ก๐ก
8 weeks: training 2 weeks: test
1/1/2016 25/2/2016 10/3/2016
13,986 users
19. 19Dataset 2
Users Artists Logs for train Sessions for train Logs for test Sessions for test
2,850 12,360 872,614 106,840 201,966 24,958
๏ฌ Publicly available Play logs of Last.fm by Schedl in ICMRโ16
๏ฌ Consists of 10 weeks logs from 1/1/2013 to 11/3/2013
๏ฌ 8 weeks are used as training data and 2 weeks as test data
โฎ
๐ก๐ก
๐ก๐ก
๐ก๐ก
8 weeks: training 2 weeks: test
1/1/2013 26/2/2013 11/3/2013
2,850 users
21. 21Results
Japanese service
Perplexity
Number of topics (K)
0
100
200
300
400
500
0 100 200 300
Zheleve
Ours
Number of topics (K)
500
800
1100
1400
1700
2000
2300
2600
0 100 200 300
Zheleve
Ours
Last.fm
๏ฌ Our model outperforms the model by Zheleve regardless of
the number of topics in both datasets
๏ฌ We did not evaluate for ๐พ๐พ > 300 because if ๐ฒ๐ฒ becomes too
large, many similar topics appear and it is difficult to carry
out the qualitative evaluation
25. 25Usersโ addiction ratio distribution
๏ฌ Users with > 0.9 addiction ratio achieve the second highest peak
๏ฌ Many users are addicted to artists to a certain extent: this result
further indicates the usefulness of considering the addiction mode
Prob
T A
0.72
Prob
T A
0.11
Prob
T A
0.53
โฎ
13,986 users
0
1500
3000
4500
6000
7500
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequencyofusers
Addiction ratio
T: taste
A: addiction
27. 0.55
0.6
0.65
0.7
0.75
0 2 4 6 8 10 12 14 16 18 20 22
27Per-hour addiction ratio
All usersโ play logs between 6 AM and 7 AM
๐๐ ๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก = 0.22
๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = 0.78
๐๐ ๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก = 0.59
๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = 0.41
๐๐ ๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก = 0.75
๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = 0.25
โฎ
Sum ๐๐(๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก) and ๐๐(๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐) of all logs respectively and normalize their sum to 1
Hour
Addictionratio
๏ฌ Addiction ratio is high in the early morning and low at night
๏ฌ This knowledge can be applied to music recommendation
(e.g., recommend unknown artists to the user at night)
High addiction
28. 28Day of week based addiction ratio
0.6
0.61
0.62
0.63
0.64
0.65
Mon Tue Wed Thu Fri Sat Sun
Addictionratio
๏ฌ Addiction ratio is high on weekdays and low on weekends
๏ฌ These results enable us to analyze peopleโs music listening
behavior from a new viewpoint
Low addiction
30. 30Per-topic addiction ratio
All usersโ play logs of top 20 artists in topic 19
Topic ID
0%
20%
40%
60%
80%
100%
10 8 19 20 1 4 26 12 21 30 2 9 13 18 11 6 14 25 5 24 15 3 22 27 23 16 29 7 28 17
addiction taste
๐๐ ๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก = 0.66
๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = 0.34
๐๐ ๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก = 0.52
๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = 0.48
๐๐ ๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก๐ก = 0.74
๐๐ ๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐๐ = 0.26
โฎ
๏ฌ Topic 8 (low addiction) is related to anime songs and
topic 28 (high addiction) is related to old Japanese artists
๏ฌ These results would be useful to generate a playlist
31. 31Summary
๏ผCarried our a quantitative evaluation
Our model outperformed the one considering only userโs taste in music
๏ผCarried out qualitative evaluations
Usersโ addiction ratio distribution, per-hour addiction ratio etc
๏ผProposed a new model for music listening behavior
Our model considers userโs usual taste in music and the addiction to artists
Contributions
Future work
๏ฌ Use the knowledge obtained from log analysis for
artist recommendation and play list generation
๏ฌ Extend our model by considering the time transition of addiction