Representing natural language causality in Conceptual Graphs: The Higher order conceptual relation problem

Abstract

This paper studies the conceptual relations expressed in NL-texts in order to represent texts meaning in Conceptual Graphs. Causality is taken as a working example. Representation is considered first. Causal connectors are defined in terms of CGs, and algorithms show how these definitions combine with the connected constituent graphs. As conceptual relations hold between situations or events rather than between simple concepts, their arguments in connector definitions are constrained to be situation concepts. After this the specific case of argumentative relations involving speech situation is examined. A simple representation of these situations is proposed and an additional constraint is introduced in argumentative connector definitions (e.g. since). A second aspect, interpretation, is also looked at so as to tackle the problem of implicit yet informative relations. A partial order on relations is used to simulate human understanding of implicit links in the light of our propensity to interpret everything in causal terms.