PhotoSlap: A Multi-player Online Game for Semantic Annotation

Multimedia content presents special challenges for the search engines, and could benefit from semantic annotation of images. Unfortunately, manual labeling is too tedious and time-consuming for humans, whereas automatic image annotation is too difficult for the computers. In this paper, we explore the power of human computation by designing a multi-player online game, PhotoSlap, to achieve the task of annotating metadata for a collection of digital photos. PhotoSlap engages users in an interactive game that capitalizes on human ability in deciphering quickly whether the same person shows up in two consecutive images presented by the computer. The game mechanism supports the objection and trap actions to encourage truthful input from the players. This research extends human computation research in two aspects: game-theoretic design principles and quantitative evaluation metrics. In particular, PhotoSlap can be shown to reach subgame perfect equilibrium with the target strategy when players are rational and without collusion. Experiments involving four focus groups have been conducted, and the preliminary results demonstrated the game to be fun and effective in annotating people metadata for photo collections.