The problem we address in this paper is the role of context in the interpretation of images when pictures are used as queries. An image usually depicts several objects and is open to a great number of interpretations. The problem is how to determine the intention of an image query. We suggest that this problem is structurally similar to the problem of how to interpret an ambiguous sentence, and that the task can be modelled in a similar way. The role of the context is a key factor in the solution of the problem of disambiguation. But we argue that existing accounts of context do not explain what role the context plays in this. It is then shown how the disambiguation of images as queries can be modelled as a game of partial information. On the basis of this, a more precise account of the role of context in communication is proposed.
Dag Elgesem, Joan C. Nordbotten