Linguistic and Computational Approaches to Information in Discourse: Theme, Focus, Given and Other Dangerous Things

Abstract

This paper addresses the information communication problem faced by interlocutors during the process of text production and understanding. Different processes govern the flow of information in discourse, such as the process of content selection, the process of information distribution, and the process of thematic organization.These processes have to be encoded by the text producer into a single linear structure to be decoded by the text receiver in the complementary process of discourse comprehension. Different disciplines, such as Linguistics and Artificial Intelligence, have studied these phenomena from different perspectives, thus developing different, but related categories, with overlapping meanings in many occasions, for the study of information structure in discourse. This paper reviews the commonalities and the different uses of categories such as topic, theme, given/new, and focus in both disciplines, and provides operational redefinitions for these notions in an attempt to integrate them into a unified account of information structure in discourse which might be useful for both descriptive and computational purposes.