Conventional speech act formulae in HPSG

The assumptions that we will make about formulae are compatible with the idea that they are conventions of usage as discussed by Morgan (1978). However, Morgan and most subsequent authors are vague on the precise characterization of convention and the empirical criteria by which something is to be classified as conventional. Our own work is based on a corpus of over 2,000 recorded spontaneous exchanges in Cypriot Greek (hence CG) described in Terkourafi (2001). This is probably the most extensive survey that has been carried out for any language where the full context of natural interactions was directly observed (as opposed to studies on previously collected corpus data or experiments with artificial tasks). We discuss Terkourafi’s data and her empirically-grounded notion of formula further in §2. There has been one previous attempt at representing this data in HPSG (Terkourafi and Villavicencio, 2003): we will discuss the differences between this and the approach suggested here below. A defining characteristic of HPSG is the assumption that it should be possible to represent semantics and pragmatics as well as syntax. There has been relatively little work on pragmatics in HPSG and, in fact, we do not wish to argue that a full account of pragmatics should be encoded within HPSG, if only because support for the sort of inferences required to interpret utterances such as (1) requires full-blown reasoning over meaning representations from utterances, non-linguistic context and general world knowledge which does not fit well into an essentially sign-based account. However, to the extent that illocutionary force is conventionally associated with particular lexemes or constructions, it should be possible to derive it within an HPSG.