1 / 46

Exploring social networks with Matrix-based representations

Exploring social networks with Matrix-based representations. Nathalie Henry. Co-supervisors: Jean-Daniel Fekete & Peter Eades. Background. Peter Eades, Univ. of Sydney. Sociologists from EHESS, Historians from the French Archives, Analysts from EDF and France telecom.

yana
Download Presentation

Exploring social networks with Matrix-based representations

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Exploring social networks with Matrix-based representations Nathalie Henry Co-supervisors: Jean-Daniel Fekete & Peter Eades Exploring social networks with matrices

  2. Background Peter Eades, Univ. of Sydney Sociologists from EHESS, Historians from the French Archives, Analysts from EDF and France telecom Jean-Daniel Fekete, INRIA Exploring social networks with matrices

  3. Social Networks • What is it? • A set of actors connected by relations • a HOT topic • Online communities (FaceBook, Flickr) • Online collaboration (Wikipedia, Sourceforge) • Scientific collaboration • And more… • The WEB • Diseases transmission networks • Terrorists networks FaceBook Interactive Graph Internet Traffic [Eick et al.] Exploring social networks with matrices

  4. Analyzing social networks Anscombe’s numbers Statistics Visual representation • Answering questions with statistics • Exploratory Data Analysis [Tukey,77] • Answering question you did not know you had • Looking at the raw data from different perspectives Exploring social networks with matrices

  5. Network Visualization • What social scientists start with: • Node-Link Diagram • Overlapping nodes • Edge-crossing  Identify connections 100+ actors - InfoVis community Exploring social networks with matrices

  6. Network Visualization [Ke et al.,04] • What social scientists end with: • Most important tasks • Communities • Central Actors  Discover the structure of the network Exploring social networks with matrices

  7. Large Network Visualization 4000+ actors • What social scientists start with: • Node-Link Diagram • Overlapping nodes • Edge-crossing  Identify connections Exploring social networks with matrices

  8. Large Network Visualization 4000+ actors • What social scientists end with: If it is LARGE, it takes a looong time… but if it is DENSE, it is not possible ! Exploring social networks with matrices

  9. What are the solutions? Sampling, filtering Clustering into meta-nodes Alternative representations such as adjacency matrices A A B C D 1 1 C 1 A 0 B 1 0 0 0 A A B C D 1 C 0 0 0 1 1 1 C A 0 D B D 0 0 0 0 B 1 1 0 0 1 C 1 1 0 D B D 1 0 1 0 A year of emails So, let’s investigate the potential of MATRICES ! Exploring social networks with matrices

  10. My Approach Observation Evaluation Brainstorming Prototyping • Participatory Design • Outcomes • They use statistics, graph analytics • They use mainly node-link diagrams • But they use sometimes matrices Exploring social networks with matrices

  11. My Approach Exploration Perception Information Visualization Communication Exploring social networks with matrices

  12. Itinerary of my PhD [Survey in progress] 3. Augment matrices 1. Make matrices usable [Henry and Fekete, IHM06&Infovis06] 2. Combine matrix+node-link 4. Merge matrix+node-link [Henry and Fekete, Interact07] Interaction technique [Elmqvist et al., CHI08] [Henry et al., Infovis07] Case study [Henry et al., IJHCI07] Exploring social networks with matrices

  13. Outline Exploration Perception 3. Augment matrices 1. Make matrices usable 2. Combine matrix+node-link 4. Merge matrix+node-link Communication Exploring social networks with matrices

  14. 1. How to make matrices usable? • J. Bertin, 1967 High-level structure? • Groups • Trends • Outliers  Make them VISUAL  REORDER their rows and columns Exploring social networks with matrices

  15. Reordering methods A B C D E F G H I J [Survey to be published] • Lot’s of methods ! • Table-based ordering methods • Graph linearization • Mine! (mixed approach) Graph Linearization Hierchical clustering of microarray data Exploring social networks with matrices

  16. Mixed approach B E H A C F D G A B C D E F G H A B C D E F G H 1 1 0 0 0 1 1 2 2 3 0 1 0 0 1 2 0 1 0 2 1 3 1 0 0 0 1 2 3 2 0 0 1 0 1 0 0 0 0 1 0 2 2 3 1 2 0 3 1 0 1 2 0 0 0 0 0 2 0 3 3 4 1 1 0 0 2 3 1 0 0 0 1 1 3 0 2 1 0 4 0 1 0 2 1 4 0 0 0 1 3 3 2 1 0 0 0 0 0 1 0 0 0 0 2 3 3 1 4 4 0 5 1 1 0 1 3 1 0 0 0 0 0 2 2 4 5 0 [Henry and Fekete, IHM’06] [Henry and Fekete, InfoVis’06] • Place actors with similar connection patterns beside each other A B C D E F G H A B C D E F G H  Add information to the adjacency matrix Exploring social networks with matrices

  17. Reordering methods There are lots of methods… … but the real question is:  What is a GOOD ordering? Exploring social networks with matrices

  18. What is a good ordering? … for performing the analysis of a social network Most important tasks • Communities: B and C • Central Actors: A Manual ordering Exploring social networks with matrices

  19. What is a good automatic ordering? • Algorithms designed from formal measures • Can we characterize them according to the visual features they produce? Exploring social networks with matrices

  20. Empirical study [Henry and Fekete, Beliv’06] • How people perceive groups? Exploring social networks with matrices

  21. Is there A good ordering? • For a specific • visual feature? • type of data? • No :( • But it saves them time to start from somewhere! • Analysts need several orderings to find a consensus in the data I focused on assisted reordering and interactions to find consensus Exploring social networks with matrices

  22. Interacting to find a consensus [Henry and Fekete, IHM’06] [Henry and Fekete, InfoVis’06] > interactive (fuzzy) clustering > interactive ordering Exploring social networks with matrices

  23. Outline Exploration Perception 3. Augment matrices 1. Make matrices usable 2. Combine matrix+node-link 4. Merge matrix+node-link Communication Exploring social networks with matrices

  24. 2. Combining best of both worlds with MatrixExplorer Exploring social networks with matrices

  25. MatrixExplorer [Henry and Fekete, IHM’06] [Henry and Fekete, InfoVis’06] • Coordinated views • Interaction to explore • Our hypothesis: • Matrix to explore • Node-link to communicate • Our observations: • Node-link for certain tasks, matrix for others • Cognitively demanding to switch back and forth • To find a consensus, they use both representations and different layouts/orderings Exploring social networks with matrices

  26. Exploring with matrix or node-link? • Social networks • Sparse (genealogical tree)  node-link • Large/Dense (a year of emails)  matrix • Why are users switching back and forth when exploring large/dense networks? Exploring social networks with matrices

  27. Matrices weakness ? ? ? A B C D 1 1 1 A 0 A A B 1 0 0 0 B C 1 C 0 0 0 C D 0 0 0 0 D B • The problem of path-related tasks[Ghoniem et al., 2005] • Always possible on matrices… • … but tedious ! Exploring social networks with matrices

  28. Outline Exploration Perception 3. Augment matrices 1. Make matrices usable 2. Combine matrix+node-link 4. Merge matrix+node-link Communication Exploring social networks with matrices

  29. 3. Augmenting matrices with MatLink Exploring social networks with matrices

  30. MatLink [Henry and Fekete, Interact’07] Best paper award • Adding static links • Adding interactive links • For larger matrices… Exploring social networks with matrices

  31. Mélange [Elmqvist, Henry, Riche and Fekete, CHI’08] • Folding the space between 2 or more points of focus • Works well for matrices and social networks tasks Exploring social networks with matrices

  32. Exploring with matrix or node-link? • Social networks • Sparse (genealogical tree)  node-link • Large/Dense (a year of emails)  matrix • Small-world networks: • A common category of social networks • Globally sparse but locally dense  node-link?  matrix? Exploring social networks with matrices

  33. Outline Exploration Perception 3. Augment matrices 1. Make matrices usable 2. Combine matrix+node-link 4. Merge matrix+node-link Communication Exploring social networks with matrices

  34. Small-world • The best representation? • What is happening • Inside communities? • Between communities? [Auber et al., InfoVis’04] Exploring social networks with matrices

  35. 4. Merging matrix and node-link with NodeTrix Exploring social networks with matrices

  36. Interactive pen tablet Exploring social networks with matrices

  37. NodeTrix [Henry et al., InfoVis’07] • Explore • Communicate Exploring social networks with matrices

  38. Actors in one or more communities • Where to place them? • Between communities? • In one or the other community? • In overlapping communities? • In a higher level community? • Why not duplicating them? • What are the effect of duplication on user understanding? Exploring social networks with matrices

  39. Node Duplication • More accurate community view • Minimizing the misleading effect by visualizing links between duplicates Exploring social networks with matrices

  40. What about evaluation? • Participatory design • Mostly informal and during the whole process • Controlled experiment • For specific tasks (low-level) on specific datasets with specific representations • Case Study • More realistic settings • Longitudinal Study… coming Exploring social networks with matrices

  41. Case Study on 20 years of 4 HCI conferences [Henry et al., IJHCI 07] • Over 5 months • Exploratory analysis of • 332 conferences • 27 000 actors • 118 000 relations • Using MatrixExplorer • Exploring large networks • Presenting the results • Then MatLink and NodeTrix Exploring social networks with matrices

  42. Communicating information InfoVis coauthorship AVI coauthorship Exploring social networks with matrices

  43. Communicating information (2) UIST coauthorship poster Exploring social networks with matrices

  44. What I have done Exploration Perception 3. Augment matrices 1. Make matrices usable 2. Combine matrix+node-link 4. Merge matrix+node-link Communication Exploring social networks with matrices

  45. Research directions • Time-related data • On social networks • Scientific collaboration (INRIA) • Group awareness (Wikipedia France) • On other data • Biological networks (Institut Pasteur) • Reflecting on user activity (MSR) • Longitudinal and qualitative studies • Logging data • New visualizations for • analyzing the data collected • reflecting it to users: for exploration and communication Exploring social networks with matrices

  46. La FIN! I’m here! INRIA in 2006 156 teams, 8400 persons • Research approach • Participatory design • Several perspectives • Research directions • Time-related data • Evaluation QUESTIONS? Exploring social networks with matrices

More Related