From descriptions to images: what reasoning in between?

Nalig (Natural Language driven Image Generator) is a system which accepts a natural language description of a scene and draws it on a screen. In this paper an overview of the NALIG reasoning activities is given, from the analysis of spatial propositions up to the actual object positioning, addressing problems as unmentioned objects instantiation, stableness analysis through a qualitative theory of statics, relation transitivity and selective object transparency. An object taxonomy has been developed to handle a set of Italian spatial relationships. Only descriptions of static scenes are interpreted by the system on the current stage of implementation.