Experiment 2012: Can we create better links by playing games?

In 2012, we officially released Veri Links and started a survey, which we distributed on Linked Data and Semantic Web mailing lists. We recorded in-game statistics and survey answers to obtain a clearer picture on the usefulness of games, in particular VeriLinks, for interlinking. The results and accompanying material for the experiments are available here:

• Experiment Report
• Survey:
• announcement blog post
• survey on VeriLinks [can still be used for feedback, but will be offline eventually]
• all questions and results as PDF
• raw survey results as CSV (with questions)
• Data:
• used link specifications
• generated links
• manually created GOLD standard
• templates used for displaying questions for the above link tasks
• In-Game Statistics (see report for explanations):
• raw link validation statistics in CSV format, columns: ID, subject, predicate, object, link task, link specification confidence value (as computed by the above link specifications), game confidence value (explained in the report), GOLD standard (1=correct,0=incorrect), number of times the link has been played, number of times the link has been judged as correct in the game, number of times the link has been judged as incorrect in the game
• Weka machine learning toolkit training data file

Remark: Since people participated in the survey and played the game outside of the 3-day experiment phase, the raw data above are updated versions of those in the report.

Survey Winners (those who agreed that we can publish their names): Gianluca Demartini, Kingsley Idehen, Joe Montibello, Michael Röder, Jakub Kotowski, Thomas Efer, Virginia Knight, Ruby, Florent Jochaud