Experiment-Driven Development of a GWAP for Marking Segments in Text
Chris Madge, Jon Chamberlain, Udo Kruschwitz, Massimo Poesio · 2017
This paper describes TileAttack, an innovative highly configurable game-with-a-purpose (GWAP) designed to gather annotations for text segmentation tasks whilst exploring the effects of different game mechanics on GWAP for NLP (Natural Language Processing) problems, with a view to improving both quality of player contributions and player uptake. In this work we present a pilot experiment that shows TileAttack labelling "mentions" and being used to test the effects of in game time constraints on accuracy and player engagement. We present the results of this experiment using a set of metrics derived from those used for evaluating Free-To-Play (F2P) games.