Focus and coherence in discourse
Leslie Graham Binet · ERA · 1985
Since Frege, many people have regarded meaning as a relation between an expression in a language and entities in the world, a relation that is essentially fixed in character, so that in a given context the expression always has the same extension. But it is not clear that this approach is really appropriate for cognitive science, and it has been suggested that what we really need is a "procedural semantics", relating utterances to mental representations rather than to the world (see, for example, Isard and Davies 1972, Johnson-Laird 1977, and Woods 1981). At the same time more attention is being paid to pragmatics and the effects of context on meaning. Both these trends are represented, for example, in Johnson-Laird's theory of mental models (Johnson-Laird 1983), in which representations of discourse content, incorporating pragmatic knowledge of the world, guide the process of semantic interpretation in such a way that sense and reference interact dialectically, as it were. In keeping with this sort of approach, I would like to suggest two methodological principles. First, that linguistic structure should always be explained in terms of its function in the cognitive process (and hence that semantic structure is best explained by something like procedural semantics). Second, that a theory of semantics must include some account of pragmatics. In other words, I would like to move away from the notion that language can be formally characterised as an independent abstract structure, and look at how it is used instead. I think computational linguistics may have something special to contribute in this area, and I shall try to illustrate this with a discussion of some discourse phenomena. This paper looks at how the theory of focus can be used to explain some aspects of discourse coherence, particularly those related to the interpretation of anaphora. Part One is a general discussion of such theories, concentrating on two theories of local focus. Part Two looks in detail at a computer model based on one of these. Part Three presents an extended example of discourse analysis based on the theory of focus, and looks at some ways in which it might be extended.