Oliver Lemon

Constraint Matching for Diagram Design: Qualitative Visual Languages

This paper examines diagrams which exploit qualitative spatial relations (QSRs) for representation. Our point of departure is the theory that such diagram systems are most effective when their formal properties match those of the domains that they represent (e.g. [1, 2, 3]). We argue that this is true in certain cases (e.g. when a user is constructing diagrammatic representations of a certain kind) but that formal properties cannot be studied in isolation from an account of the cognitive capacities of diagram users to detect and categorize diagram objects and relations. We discuss a cognitively salient repertoire of elements in qualitative visual languages, which is different from the set of primitives in mathematical topology, and explore how this repertoire affects the expressivity of the languages in terms of their vocabulary and the possible spatial relations between diagram elements. We then give a detailed analysis of the formal properties of relations between the diagram elements. It is shown that the analysis can be exploited systematically for the purposes of designing a diagram system and analysing expressivity. We demonstrate this methodology with reference to several domains, e.g. diagrams for file systems and set theory (see e.g. [4]).

Probabilistic Dialogue Modelling

We show how Bayesian networks and related probabilistic methods provide an efficient way of capturing the complex balancing of different factors that determine interpretation and generation in dialogue. As a case study, we show how a probabilistic approach can be used to model anaphora resolution in dialogue.