The Role of Conceptual and Linguistic Ontologies in Discourse

In this paper we argue that a clear division between two sources of information— one oriented to world knowledge, the other to linguistic semantics—offers a framework within which mechanisms for modelling the highly flexible relation between language and interpretation necessary for natural discourse can be specified and empirically validated. Moreover, importing techniques and results from formal ontology provides the formal underpinnings necessary for representing these sources of information appropriately. The result is a computationally specifiable model of dialogic interaction within which flexible discourse interpretation is seen as the result of inter-ontology mappings between our two sources of information. These mappings are dynamically negotiated according to the concrete discourse moves of interlocutors. The paper draws on our ongoing empirical studies in spatial discourse, where interlocutors jointly solve varieties of spatially-embedded tasks, and suggests that ontological formalization is of considerable benefit for constructing computational dialogue systems.