Exploring the Role of Punctuation in the Signalling of Discourse Structure

This paper reports on the beginnings of some work whose aim is to investigate the role of punctuation in signalling discourse structure. This is part of a larger enterprise motivated by the view that natural language processing systems, whether generating or understanding language, should take account of the physical manifestation of language: in the context of written language, this means taking account of the constraints imposed and opportunities offered by the real estate of the pages (or whatever other surfaces) on which the words appear. In this context, punctuation marks occupy an interesting middle ground between overtly linguistic phenomena like cue words and phrases, and overtly layout-oriented phenomena such as indentation and spacing. Ultimately, we need some means of representation that can accommodate both these dimensions of language use. This paper suggests that a study of the uses of punctuation can be a useful first step towards identifying what the elements of such a representation language should be.