A COMPUTATIONAL FRAMEWORK FOR UNIFYING VISION AND LANGUAGE

This paper presents a knowledge representation framework that provides an integrated approach to reasoning about vision and language in spat ial domains. Images are stored using a representation that preserves i nformation such as shape and distance; linguistic information is repre sented in a descriptive, semantic network. We incorporate a model-base d formalism as an intermediate representation, which can be used to tr ansform visual to descriptive representations and vice versa.