With each device or application that expands the bandwidth of available information, the computer...

Preview:

Citation preview

With each device or application that expands the bandwidth of available

information, the computer’s understanding of us remains

unchanged.

2

Social Network Analyses about Web Services

3

Examples

Hyperlink structure between personal homepages(Adamic and Adar, 2003)

Discussion relationship in BBS(Goh et al., 2006)

Recommendation networks in Amazon(Leskovec et al., 2007)

Interaction Patterns in Yahoo Answers(Adamic et al., 2008)

Friendship and followship in Twitter(Huberman et al., 2009)

Community structure in Facebook(Traud et al., 2011)

And so on, and so forth…

4

Friendship Networks

5

Friends and neighbors on the Web(Adamic and Adar, 2003)

Data: Students’ homepages at (a) Stanford and (b) MIT

Result:

6

Adamic and Adar (2003) (2/2)

Summary of links given and received among personal homepages:

PowerLaw

7

Find Me If You Can: Improving Geographical Predictionwith Social and Spatial Proximity (Backstrom et al., 2010)

Data: Fackbook

Result:

8

Backstrom et al. (2010) (2/2)

9

Social networks that matter:Twitter under the microscope (Huberman et al., 2009)

Data: Twitter

Result:

10

Huberman et al. (2009) (2/4)

11

Reciprocal friends

Huberman et al. (2009) (3/4)

12

Huberman et al. (2009) (4/4)

It’s friends that matter

13

Friendship networks and social status(Ball and Newman, 2012)

Data: Friendships among students at US high and junior high schools

Result:

14

Ball and Newman (2012) (2/3)

15

Ball and Newman (2012) (3/3)

16

Online Discussion Networks

17

Structure and evolution of online social relationships: Heterogeneity in unrestricted discussions (Goh et al., 2006)

Data: A University BBS

Result: Schematic network snapshots of

a) the BBS network

b) traditional social network

18

Visualizing the Signatures of Social Rolesin Online Discussion Groups (Welser et al., 2007)

Data: Usenet newsgroups

Result: answer person vs. discussion person

19

Welser et al. (2007) (2/7)

20

Welser et al. (2007) (3/7)

21

Welser et al. (2007) (4/7)

22

Welser et al. (2007) (5/7)

23

Welser et al. (2007) (6/7)

24

Welser et al. (2007) (7/7)

25

Community Structure and Information Flow in Usenet:Improving Analysis with a Thread Ownership Model (McGlohon and Hurst, 2007)

Data: Political newsgroups of Usenet

Result: Cross-posting network

26

McGlohon and Hurst (2007) (2/4)

Anomalies: The points far below the fitting line (with abnormally low reply rat

es) are tw domains. The ones above the fitting line (high reply rates) tend to be in Euro

pean domains.

27

McGlohon and Hurst (2007) (3/4)

The most reciprocated group (hun.politika) had a reciprocity of up to 0.58, and the least reciprocated group tw.bbs.soc.politics, had a reciprocity of 0.057.

The low-reciprocity groups generally had low traffic (fewer than 100 authors in any given year, with the exception of tw.bbs.soc.politics).

All of Taiwan-based groups in our data had very low reciprocity.

28

McGlohon and Hurst (2007) (4/4)

Post ownership ratio: fr.soc.politique has a ratio of 0.92 tw.bbs.soc.politics.kmt’s was around 0.003

29

Expertise Networks in Online Communities:Structure and Algorithms (Zhang et al., 2007)

Data: The Java Forum, a large online help-seeking community

Result:

30

Zhang et al. (2007) (2/2)

In the Java Forum, there are some extremely active users who answer a lot of questions while a majority of users answer only a few. (See in degree)

Likewise, many users ask only a single question, but some ask a dozen or more. (See out degree)

31

Knowledge Sharing and Yahoo Answers:Everyone Knows Something (Adamic et al., 2008)

Data: Yahoo Answers (YA), a large and diverse question-answer forum

Result:

32

Adamic et al. (2008) (2/4)

33

Adamic et al. (2008) (3/4)

34

Adamic et al. (2008) (4/4)

35

Online Recommendation Networks

36

The Dynamics of Viral Marketing(Leskovec et al., 2007)

Data: a person-to-person recommendation network, consisting of 4 milli

on people who made 16 million recommendations on half a million products (Amazon?)

Result:

37

Leskovec et al. (2007) (2/3)

38

Leskovec et al. (2007) (3/3)

39

Leskovec et al. (2007) (3/3)

40

Social Influence and the Diffusion of User-Created Content (Bakshy et al., 2010)

Data: Second Life, a massively multiplayer virtual world

Result:

Most assets in the data set are ownedby a relatively small number of users,and very large assets of size 1,000 orgreater make up less than 10% ofall assets.

This is the familiar long tailof content popularity.

41

Bakshy et al. (2010) (2/6)

42

Bakshy et al. (2010) (3/6)

popularity

transfersbetweenfriends

transfersthat

result intransfers

43

Bakshy et al. (2010) (4/6)

44

Bakshy et al. (2010) (5/6)

45

Bakshy et al. (2010) (6/6)

46

Information Propagation Networks

47

How to search a social network(Adamic and Adar, 2005)

Data: a network of actual email contacts within HP Labs

Result:

48

Adamic and Adar (2005) (2/4)

49

Adamic andAdar (2005) (3/4)

50

Adamic and Adar (2005)(4/4)

Probability of two individualscorresponding by email as a function of

the distance between their cubicles

Email communicationswithin HP Labsmapped onto

approximate physical location

51

Organizational chart and advice networkin a business unit (Krackhardt, 1996)

52

(Krackhardt, 1996) (2/2)

53

A Measurement-driven Analysis ofInformation Propagation in the Flickr Social Network (Cha et al., 2009)

Data: Flickr

Result:

54

Cha et al. (2009) (2/3)

55

Cha et al. (2009) (3/3)