1. Chapter 1. Social Media and Social Computing October 2012 Youn-Hee Han http://link.koreatech.ac.kr

2. 1.1 Social Media  A rapid development and change of the Web and the Internet –Participatory web application and social networking sites Empowering them with new forms of collaboration Communication –Wikipedia Much numbers of online volunteers collaboratively write encyclopedia articles –Amazon (Online Market) and Social Commerce They recommend products by tapping on crowd wisdom via user shopping and reviewing interactions; –Twitter Political movements benefit from new forms of engagement and collective actions –Facebook Connecting People 2

3. 1.1 Social Media  Facebook – Big Change of Our Life –901 million monthly active users at the end of March 2012. –More than 125 billion friend connections on Facebook at the end of March 2012. 3

4. 1.1 Social Media  Classical web and traditional media –1 : N  Present social media –N : M 4

5. 1.1 Social Media  A user of social media can be both a consumer and a producer.  This new type of mass publication enables the production of timely news and grassroots ( 일반인들에 의한 ) information and leads to mountains of user-generated contents, forming the wisdom of crowds. (Collective Intelligence)  Distinctive characteristic of social media –Participation –Sharing –Rich user interaction 5

6. 1.2 Concepts and Definitions  Social Networks –A social network is a social structure made of nodes (individuals or organizations) and edges that connect nodes in various relationships (or interdependencies) like friendship, kinship, etc.  Why Social Network in Research Community? –All entities (e.g., people, devices, or systems) in this world are related to each other in one way or another –It can be used in the context of information and communication technologies to provide efficient data exchange, sharing, and delivery services –By using a social network, we can use the knowledge about the relationship to improve efficiency and effectiveness of network services 6

7. 1.2 Concepts and Definitions  Networks and Representations –Graphical representation, Matrix representation –In a weighted network, edges are associated with numerical values. –In a signed network, some edges are associated with positive relationships, some others might be negative. –Directed networks have directions associated with edges. In our example in Figure 1.1, the network is undirected. Figure 1.1 7

8. 1.2 Concepts and Definitions  Networks and Representations –Example of Directed Social Networks: Twitter one user x follows another user y, but user y does not necessarily follow user x In this case, the follower-followee network is directed and asymmetrical 8

9. 1.2 Concepts and Definitions Figure 1.1 9

10. 1.2 Concepts and Definitions  Nomenclature ( 용어 체계 ) –The radius of a network is the minimum eccentricity among the vertices of the network () radius(G)=3 –The diameter of a network is maximum eccentricity among the vertices of the network (i.e., the length of the longest geodesic) () diameter(G)=5 –The center of a network is the set of vertices of eccentricity equal to the radius () Center(G)={4, 5, 6} –The periphery of a network is the set of vertices of eccentricity equal to the diameter () Center(G)={2, 9} Figure 1.1 10

11. 1.2 Concepts and Definitions  Properties of large-scale networks –Networks in social media are often very huge, with millions of actors and connections. –These large-scale networks share some common patterns scale-free distributions small-world effect strong community structure. –Simple Networks a lattice graph or random graphs. –Complex Networks Networks with non-trivial topological features are called complex networks to differentiate them from simple networks 11

12. 1.2 Concepts and Definitions  Power law distribution –Node degrees in a large-scale network often follow a power law distribution Most nodes have a low degree, while few have an extremely high degree (say, degree > 10 4 ) Low degree Long tail 12

13. 1.2 Concepts and Definitions  Scale-free distribution –Such a pattern is also called scale-free distribution the shape of the distribution does not change with scale. if we zoom into the tail (say, examine those nodes with degree > 100), we will still see a power law distribution –This self-similarity is independent of scales. –Networks with a power law distribution for node degrees are called scale-free networks 13

14. 1.2 Concepts and Definitions  Small-world effect –Travers and Milgram (1969) conducted an experiment to examine the average path length for social networks of people in the United States “six degrees of separation” –Leskovec and Horvitz (Microsoft, 2008) This result is also confirmed recently in a planetary-scale instant messaging network of more than 180 million people, in which the average path length of any two people is 6.6 Washington Post Article  http://www.washingtonpost.com/wp- dyn/content/article/2008/08/01/AR2008080103718.html?nav=hcmodule http://www.washingtonpost.com/wp- dyn/content/article/2008/08/01/AR2008080103718.html?nav=hcmodule –Most real-world large-scale networks observe a small diameter 14

15. 1.2 Concepts and Definitions 15

16. 1.3 Challenges  Flood of data allows for an unprecedented large-scale social network (complex networks) analysis –millions of actors or even more in one network. email communication networks, instant messaging networks, mobile call networks, friendship networks, co-authorship or citation networks, biological networks, metabolic pathways, genetic regulatory networks and food web.  These large-scale networks present novel challenges for mining social media.  Some examples are given below: 16

17. 1.3 Challenges  Scalability. –Networks of this astronomical size!  Heterogeneity. –Two persons can be friends and colleagues at the same time.  Evolution. –Social media emphasizes timeliness.  Collective Intelligence. –Wisdom of crowds.  Evaluation –A research barrier concerning mining social media is evaluation. 17

18. 1.4 Social Computing Tasks  Network Modeling –Since the seminal work by Watts and Strogatz (1998), and Barabási and Albert (1999), network modeling has gained some significant momentum. –Researchers have observed that large-scale networks across different domains follow similar patterns, such as scale-free distributions, the small-world effect and strong community structures as we discussed in Section 1.2.2. YoutubeFlickr 18

19. 1.4 Social Computing Tasks  Network Modeling –When networks scale to over millions and more nodes, it becomes a challenge to compute some network statistics such as the diameter and average clustering coefficient. One way to approach the problem is sampling. Others explore I/O efficient computation. Recently, techniques of harnessing the power of distributed computing are attracting increasing attention. 19

20. 1.4 Social Computing Tasks  Centrality analysis –It identifies the most “important” nodes in a network (Wasserman and Faust, 1994). degree centrality betweenness centrality closeness centrality eigenvector centrality  equivalent to Pagerank scores (Page et al., 1999)  Influence modeling –It aims to understand the process of influence or information diffusion. –Researchers study how information is propagated (Kempe et al.,2003) and how to find a subset of nodes that maximize influence in a population. 20

21. 1.4 Social Computing Tasks  Community Detection –Community Groups, clusters, cohesive subgroups, modules in different contexts. –It is one of the fundamental tasks in social network analysis. –The founders of sociology claimed that the causes of social phenomena were to be found by studying groups rather than individuals (Hechter (1988), Chapter 2, Page 15). 21

22. 1.4 Social Computing Tasks  Community Detection –Recent Community Detection Research Scaling up community detection methods to handle networks of colossal sizes. Deals with networks of heterogeneous entities and interactions  Youtube »Entities (nodes): users, videos, tags »Edges: connecting to a friend, leaving a comment, sending a message Considers the temporal development of social media networks.  Facebook has grown from 14 million in 2005 to 500 million as in 2010.  As a network evolves, we can study how communities are kept abreast with its growth and evolution, what temporal interaction patterns are there, and how these patterns can help identify communities. 22

23. 1.4 Social Computing Tasks  Classification and Recommendation –A successful social media site often requires a sufficiently large population –Personalized recommendations can help enhance user experience. Classification can help recommendation.  E.g., in Facebook 23

24. 1.4 Social Computing Tasks  Classification and Recommendation –For instance, given a social network and some user information (interests, preferences, or behaviors), we can infer the information of other users within the same network. The classification task here is to know whether an actor is a smoker or a non-smoker (indicated by + and −, respectively). 24

25. 1.4 Social Computing Tasks  Privacy, Spam and Security –Privacy Many social media sites (e.g., Facebook, Google Buzz) often find themselves as the subjects in heated debates about user privacy. –Spam and Attacks Another issue that causes grave concerns in social media In blogosphere, spam blogs (a.k.a., splogs) (Kolari et al., 2006a,b) and spam comments have cropped up.  These spams typically contain links to other sites that are often disputable or otherwise irrelevant to the indicated content or context. Some spammers use fake identifiers to obtain other user’s private information on social networking sites. –Research is needed for “secure social computing platform” it is critical in turning social media sites into a successful marketplace 25

26. 1.5 Summary  Social media mining is a young and vibrant field with many promises.  Social media has kept surprising us with its novel forms and variety. –  Social media is increasingly blended into the physical world with recent mobile technologies and smart phones. 26

27. Appendix 27

28. Networks Regular Networks ( 출처 ) ; http://geza.kzoo.edu/~csardi/module/html/regular.html 1.Rings 1.A ring is a connected graph in which each vertex is connected to exactly two other vertices. 2.Lattices 1.A lattice is a graph in which the vertices are placed on a grid and the neighboring vertices are connected by an edge. A one dimensional lattice is like a ring, only it is not circular, the circle is not closed. A two dimensional lattice can be seen in the following picture : RingLattice 28

29. Tree Full Graph Edge 개수 : v·(v-1)⁄2 v ; vertices 개수 3. Trees A tree is a connected graph which contains no circles (cycles). A tree graph is usually plotted “tree-like” with its root on the top and then its branches going downward. (Hence its name.) The top vertex is called the “root” and the vertices at the next lower level are called the children of the root. In general the neighbors of a vertex at a lower level are called the children of that vertex 4. Stars A star graph is a special tree, where every vertex is connected to the root. 5. Full graph In a full graph every possible edge is realized, ie. there is an edge between every pair of vertices. Regular Graph 란 각 vertices 들을 연결하는 edge 들의 모양 (structure, topology) 이 전체 그래프에 걸쳐 계속하여 반복적으로 나타나는 형태의 그래프 29

30. Erdõs-Rényi random graphs G(n,p) graphs are generated this way: the graph contains n vertices. Then for every pair of vertices with probability p an edge is drawn connecting them. Below is a G(n,p) graph with n=100 and p=2/100. 30

31. NE MA Small world phenomenon: Milgram’s experiment [Instructions] Given a target individual (stockbroker in Boston), pass the message to a person you correspond with who is “closest” to the target. NE: Nebraska 주 MA: Massachusetts 주 [Outcome] 20% of initiated chains reached target average chain length = 6.5 “Six degrees of separation” 31

32. 규칙적제 멋대로, 무작위 Collective dynamics of “ small-world ” networks Duncan J. Watts & Steven H. Strogatz ( http://www.tam.cornell.edu/tam/cms/manage/upload/SS_nature_smallworld.pdf) 32

33. Structural metrics: Average path length 33

34. Structural Metrics: Degree distribution(connectivity) 34

35. Structural Metrics: Clustering coefficient 35

36. Regular networks – fully connected 36

37. Regular networks – Lattice 37

38. Regular networks – Lattice: ring world 38

39. Random Networks k=3 39

40. Random Networks 40

41. Small-world networks 41

42. Small-world networks 42

43. Small-world networks 43

44. Small-world networks 44

45. Scale-free networks 45

46. Scale-free networks 46

47. Scale-free networks 47

48. Scale-free networks 48

49. Scale-free networks 49

50. Scale-free networks 50

51. Case studies - Internet 51

52. Case studies - Internet 52

53. Case studies - Internet 53

54. Case studies - World Wide Web 54

55. Case studies - World Wide Web 55

56. Case studies - World Wide Web 56

57. Case studies - Actors 57