Representing discourse coherence

Abstract

We present a set of discourse structure relations that are easy to code, and develop criteria for an appropriate data structure for representing these relations.Discourse structure here refers to informational relations that hold between sentences in a discourse (cf. Hobbs, 1985). We evaluated whether trees are a descriptively adequate data structure for representing coherence. Trees are widely assumed as a data structure for representing coherence but we found that more powerful data structures are needed: In coherence structures of naturally occurring texts, we found many different kinds of crossed dependencies, as well as many nodes with multiple parents. The claims are supported by statistical results from a database of 135 texts from the Wall Street Journal and the AP Newswire that were hand-annotated with coherence relations, based on the annotation schema presented in this paper.

A PDTB-styled end-to-end discourse parser

“…In natural language processing (NLP), discourse parsing is the process of understanding the internal structure of a text and identifying the discourse relations in between its text units. Over the last three decades, researchers have proposed a number of discourse frameworks from different perspectives for the purpose of discourse analysis and parsing (Grosz and Sidner , 1986; Mann and Thompson , 1988; Polanyi , 1988; Hobbs , 1990; Lascarides and Asher , 1993; Knott and Sanders , 1998; Webber , 2004; Wolf and Gibson , 2005). However, designing and constructing such a discourse parser has been a difficult task, partially attributable to the lack of large-scale annotated data sets.…”

Recursion in Pragmatics

“…For another kind of discourse, namely written texts,Wolf and Gibson (2005) show that some 12.5% of coherence relations have a cross-serial dependency, making tree structures for discourse relations impractical.…”

Annotation for and Robust Parsing of Discourse Structure on Unrestricted Texts

“…sdrt also allows for crossing dependencies. There is evidence both within the sdrt framework and without that tree structures alone are not sufficient to adequately represent discourse structure (Wolf & Gibson 2005). Below we will present data from our corpus concerning verbs of saying that we think provide further evidence against the use of binary tree structures.…”