Fabelier Gephi Workshop {Events}

FabelierGephiWorkshop by by Sébastien Heymann, 12/01/2011 at 6pm

#Fabelier on Twitter, identi.ca, TwitterStreamGraphs, Collecta, flickr

this very wiki with Yifan Hu layout (large svg file)
note that the set on the upper left is the wiki documentation

Attended activities

• Presentation
• Workshop

Presentation

• own introduction
  • previously discussed with Mathieu Jacomy (early 2009) after the presentation at ScPo of WebAtlas.fr
    • also knowing UTC startup linkfluence.net (RTGI group)
• opening quote by Hal Varian
  • cf InformationRules
• discover patterns
  • non linear process
• mention of Ben Fry (2004 thesis)
  • Processing
• http://wiki.gephi.org/index.php/Manifesto
• architecture (Java NetBeans)
• layouts
  • Wikipedia:Force-based algorithms (graph drawing)
  • already a bit familiar through Graphviz
• metrics/statistics
  • including page rank
    • which can then be used after in ranking
• filters
  • can be combined, especially useful for boolean operators
    • e.g. not "this" in node label

Back to the Menu

Workshop

• graphformatexporter.php in code repository PmWiki graph format exporter done through GraphCSV action (and GraphVAN action)
  • Path:/pub/nodes.csv (700) Path:/pub/edges.csv (3000)
    • Path:/pub/graph.van
  • Path:/pub/FB_Wiki.gephi (with stats, ranking, layout)
  • result in
• self discovered
  • additional layouts
    • http://gephi.org/nbm-center-category/Layouts/
  • display only nodes after a filter
    • e.g. only Id with "AutoDebate.*" use regex
  • apply ranking
    • node size based on size of the page
    • node color based on number of revisions of the page
  • apply partition
    • e.g. coloring based on number of revisions
      • interesting by showing that most pages are edited less than 5 times
• own needs
  • how to zoom without a mouse
    • optional bottom right button
  • how to use the "play" feature? (seems to be called "Dynamics")
    • for wiki revisions
• crash
  • display labels

then had to leave... :'( Back to the Menu

Overall remarks and conclusions

• glad I prepared before hand
• should have checked the timetable and see that most of the things I knew were at the beginning
• sees tricky to make headless equivalent, heavy as requiring JVM

Back to the Menu

Other reviews or coverage

• here

To do

1. consider other own datasets
   1. browsing habits
      1. e.g. Firefox SQLite logs
         1. done via shell_scripts/browser_queries
      2. HTTP Graph -- A Gephi plugin analyze the web as you browse
      3. see Wandering Wikipedia: Datamining My Firefox History by Adam Fletcher, The Simple Logic
   2. Visualization
   3. Person
      1. add and link logs, e.g. edge weight = number of shared links
   4. bibliometry
      1. Seedea:Research/Bibliography
      2. Path:/ with all links overall, cf earlier exporter done for GM script
      3. also proposed in Epistemetrics
2. links to explore
   1. http://infosthetics.com/archives/2010/07/review_gephi_graph_exploration_software.html
   2. http://www.youtube.com/user/GephiOrg
   3. http://vimeo.com/channels/gephi
   4. http://ateliercartographie.com/
   5. discovered before
3. share the PmWiki recipe on http://PmWiki.org
   1. then share also on http://netwiki.amath.unc.edu/ since it is based on PmWiki
4. see also http://cgi.cse.unsw.edu.au/~wyos/skyrails/ in 3D but without layout feature
   1. using Path:/pub/graph.van
5. useful tools
   1. Gephi an open source graph visualization and manipulation software
6. do another visualization but without generated indexes
   1. thus removing *.*RecentChanges esp. Site.AllRecentChanges