A Dependency Perspective on RST Discourse Parsing and Evaluation

Abstract

Discourse structures in Rhetorical Structure Theory (RST) have classically been presented as constituency trees; and as a result, the RST discourse parsing community has largely borrowed from the syntactic constituency parsing community. The standard evaluation procedure for RST discourse parsers is a simplified variant of PARSEVAL, and most RST discourse parsers use
techniques that originated in syntactic constituency parsing. In this paper, we isolate a number of conceptual and computational problems with the constituency hypothesis. We then examine the consequences, for the implementation and evaluation of RST discourse parsers, of adopting a dependency perspective on RST structures. We analyze RST discourse parsing as dependency parsing by adapting to RST a recent proposal in syntactic parsing that relies on head-ordered
dependency trees, a representation isomorphic to headed constituency trees. We convert to head-ordered dependency trees the original trees from the RST-DT and their binarized version used by all existing RST parsers. On both versions of the RST-DT, we train a simple dependency parser and compare its performance to state-of-the-art RST discourse parsers on constituency and dependency metrics. We thus propose an evaluation framework to compare extant approaches
easily and uniformly, which the RST parsing community has lacked up to now.
The results of our experiments indicate that our simple dependency-based RST parser obtains competitive scores both on the standard constituency evaluation and the dependency metrics, although it performs better on the unlabelled structure than on predicting relations.