Slides from IEEE-THEMES 2010 colocated with ICASSP. Paper to appear in August issue of Select Topics in Signal Processing. Abstract: This paper presents an extensive analysis of a sample of a social network of musicians. The network sample is first analyzed using standard complex network techniques to verify that it has similar properties to other web-derived complex networks. Content-based pairwise dissimilarity values between the musical data associated with the network sample are computed, and the relationship between those content-based distances and distances from network theory explored. Following this exploration, hybrid graphs and distance measures are constructed, and used to examine the community structure of the artist network. Finally, results of these investigations are presented and considered in the light of recommendation and discovery applications with these hybrid measures as their basis.

1. Analysis and Exploitation of
Musician Social Networks
for Recommendation and Discovery
Ben Fields Kurt
b.fields@gold.ac.uk Jacobson
Christophe Mark Michael
Rhodes Sandler Casey

2. overview
– motivation
– dataset
– experiments
– social radio
2 Fields et. al - Analysis and Exploitation of Musician Social Networks

3. motivation
3 Fields et. al - Analysis and Exploitation of Musician Social Networks

4. motivation
Novelty Curves
4 Fields et. al - Analysis and Exploitation of Musician Social Networks

5. motivation
The Web
5 Fields et. al - Analysis and Exploitation of Musician Social Networks

6. So much music,
so little time.
6 Fields et. al - Analysis and Exploitation of Musician Social Networks

7. So much music,
so little of it good.
7 Fields et. al - Analysis and Exploitation of Musician Social Networks

8. How do we discover
good music?
8 Fields et. al - Analysis and Exploitation of Musician Social Networks

9. listening
9 Fields et. al - Analysis and Exploitation of Musician Social Networks

10. social listening
10 Fields et. al - Analysis and Exploitation of Musician Social Networks

11. social listening
11 Fields et. al - Analysis and Exploitation of Musician Social Networks

12. social listening
12 Fields et. al - Analysis and Exploitation of Musician Social Networks

13. social listening
13 Fields et. al - Analysis and Exploitation of Musician Social Networks

14. social listening
14 Fields et. al - Analysis and Exploitation of Musician Social Networks

15. dataset
15 Fields et. al - Analysis and Exploitation of Musician Social Networks

16. dataset
Sampling Myspace
Randomly
Selected Artist
16 Fields et. al - Analysis and Exploitation of Musician Social Networks

17. dataset
Sampling Myspace
Selected
Artist's top
friend
Selected Selected
Artist's top Artist's top
friend friend
Randomly
Selected Artist
Selected Selected
Artist's top Artist's top
friend friend
17 Fields et. al - Analysis and Exploitation of Musician Social Networks

18. dataset
Sampling Myspace
Artist's Artist's Artist's
top friend top friend top friend
Artist's
Artist's top friend
top friend Artist's
Selected top friend
Artist's
Artist's
top friend
top friend
Selected Selected Artist's
Artist's Artist's top friend
Artist's top friend top friend
top friend
Randomly
Artist's
Selected
top friend
Artist
Artist's
top friend
Selected Selected
Artist's
Artist's Artist's
top friend
top friend top friend
Artist's
top friend
Artist's
Artist's top friend
Artist's
top friend
Artist's top friend
top friend
Artist's
top friend
18 Fields et. al - Analysis and Exploitation of Musician Social Networks

19. dataset
Sampling Myspace
– scale-free (mostly)
– 15,478 nodes (artist pages)
– 120,487 directed edges
– 91,326 undirected edges
– avg. degree
– 15.5 as a directed graph
– 11.8 when undirected
19 Fields et. al - Analysis and Exploitation of Musician Social Networks

20. dataset
Cumulative Degree Distribution
20 Fields et. al - Analysis and Exploitation of Musician Social Networks

21. dataset
Cumulative Degree Distribution
21 Fields et. al - Analysis and Exploitation of Musician Social Networks

22. experiments
22 Fields et. al - Analysis and Exploitation of Musician Social Networks

23. experiments
Geodesic v. Acoustic Distance
–pair nodes by geodesic distance
–looking for correlation with
pairwise EMD
–result is inconclusive
23 Fields et. al - Analysis and Exploitation of Musician Social Networks

24. experiments
Geodesic v. Acoustic Distance
24 Fields et. al - Analysis and Exploitation of Musician Social Networks

25. experiments
Max Flow v. Acoustic Distance
– pairs of artist nodes grouped based on
Maximum Flow
– a randomized network was created as well to
compare the relationship
– results point toward a mostly orthogonal
relationship
– examining the mutual information shows that
most information not common across spaces
25 Fields et. al - Analysis and Exploitation of Musician Social Networks

26. experiments
Max Flow v. Acoustic Distance
26 Fields et. al - Analysis and Exploitation of Musician Social Networks

27. experiments
Max Flow v. EMD
27 Fields et. al - Analysis and Exploitation of Musician Social Networks

28. experiments
Max Flow v. marsyas distance
28 Fields et. al - Analysis and Exploitation of Musician Social Networks

29. experiments
Low Entropy Communities
–looking at whether communities
are more homogenous if edges
are weighted with sonic
similarity
–uses genre entropy
29 Fields et. al - Analysis and Exploitation of Musician Social Networks

30. experiments
Low Entropy Communities
algorithm c SC Srand Q
none 1 1.16 - -
gm 42 0.81 1.13 0.61
gm+a 33 0.90 1.13 0.64
wt 195 0.80 1.08 0.61
wt+a 271 0.70 1.06 0.62
Table 1. Results of the community detection algorithms
where c is the number of communities detected, SC is the
average genre entropy for all communities, Srand is the
average genre entropy for a random partition of the network
30 Fields et. al - Analysis and Exploitation of Musician Social of communities, and Q is the modu-
into an equal number Networks

31. experiments
Low Entropy Communities
algorithm c SC Srand Q
none 1 1.16 - -
gm 42 0.81 1.13 0.61
gm+a 33 0.90 1.13 0.64
wt 195 0.80 1.08 0.61
wt+a 271 0.70 1.06 0.62
Table 1. Results of the community detection algorithms
where c is the number of communities detected, SC is the
average genre entropy for all communities, Srand is the
average genre entropy for a random partition of the network
31 Fields et. al - Analysis and Exploitation of Musician Social of communities, and Q is the modu-
into an equal number Networks

32. social radio
32 Fields et. al - Analysis and Exploitation of Musician Social Networks

33. social radio
Weighted Max Flow Playlists
–max flow presents an interesting
opportunity to create playlists
using least resistant paths
–preliminary testing shows promise
–needs more exhaustive testing
33 Fields et. al - Analysis and Exploitation of Musician Social Networks

34. social radio
Playlist Generator
34 Fields et. al - Analysis and Exploitation of Musician Social Networks

35. social radio
Playlist Generator
34 Fields et. al - Analysis and Exploitation of Musician Social Networks

36. social radio
The Social Radio
– produce playlists via weighted
distance paths
– next destination song is determined
via a vote across all listeners
– candidate songs selected from
disparate communities
35 Fields et. al - Analysis and Exploitation of Musician Social Networks

37. social radio
The Social Radio
36 Fields et. al - Analysis and Exploitation of Musician Social Networks

38. resources
– http://mypyspace.sourceforge.net/
– http://dbtune.org/myspace/
– http://omras2.doc.gold.ac.uk/software/fftExtract/
– slides: http://slideshare.com/BenFields
– contact: b.fields@gold.ac.uk
http://blog.benfields.net
twitter: @alsothings
37 Fields et. al - Analysis and Exploitation of Musician Social Networks

39. resources
– http://mypyspace.sourceforge.net/
– http://dbtune.org/myspace/
– http://omras2.doc.gold.ac.uk/software/fftExtract/
– slides: http://slideshare.com/BenFields
– contact: b.fields@gold.ac.uk
http://blog.benfields.net
twitter: @alsothings
Questions?
37 Fields et. al - Analysis and Exploitation of Musician Social Networks