Identifying Congruent Pragmatic Relations in Procedural Texts

Abstract

In this paper, we present a methodology for the contrastive analysis of comparable corpora of instructional texts in different languages. The methodology is insensitive to the fact that the texts under comparison differ widely in their semantic content, and it can be reliably applied by multiple analysts. We show the results of an empirical study of cross-linguistic variation between Portuguese, French, and English instructions which follows this methodology. Using consumer instructions for ordinary household products in the three languages, we examine expressions of the two semantic relations, generation and enablement (cf. Goldman, 1970), and their available surface syntactic expressions. We examine the role of discourse perspective, as realised by rhetorical relations such as those employed within the framework of Rhetorical Structure Theory (RST), in further narrowing down the range of choices. We demonstrate that the three languages of study tolerate different levels of ambiguity, and prefer different forms of disambiguation and pragmatic signalling, attesting to the value of empirical methods for contrastive discourse study. The analysis was conducted with the aim of informing all levels of decision, from meaning to surface syntax, in the automatic generation of sets of instructional texts in those languages.