Modelling the Usage of Discourse Connectives as Rational Speech Acts

Abstract

Discourse relations can either be implicit or explicitly expressed by markers, such as 'therefore' and 'but'. How a speaker makes this choice is a question that is not well understood. We propose a psycholinguistic model that predicts whether a speaker will produce an explicit marker given the discourse relation s/he wishes to express. Based on the framework of the Rational Speech Acts model, we quantify the utility of producing a marker based on the information-theoretic measure of surprisal, the cost of production, and a bias to maintain uniform information density throughout the utterance. Experiments based on the Penn Discourse Treebank show that our approach outperforms stateof-the-art approaches, while giving an explanatory account of the speaker's choice.

Limits to the Rational Production of Discourse Connectives

“…While the uniform information density hypothesis is most suitable for studying phenomena where the production variants are meaning-equivalent, the rational speech act theory also involves reasoning about alternative meanings of an utterance, and hence seems best suited for studying the production of discourse connectives. In fact, the RSA model has already been used to account for the distribution of explicit and implicit discourse connectives, and found it to be in line with the qualitative prediction of the RSA model (Yung et al, 2016(Yung et al, , 2017. They found that, in the Penn Discourse Treebank (Rashmi et al, 2008), an explicit connective is more often omitted when the it is not informative enough to offset its production cost, or if there are enough other discourse signals in the arguments.…”