A Novel Human Computation Game for Critique Aggregation

We present a human computation game based on the popular board game - Dixit. We ask the players not only for annotations, but for a direct critique of the result of an automated system.We present the results of the initial run of the game, in which the answers of 15 players were used to profile the mistakes of an aspect-based opinion mining system. We show that the gameplay allowed us to identify the major faults of the extracted opinions. The players' actions thus helped improve the opinion extraction algorithm.