A Simple Global Neural Discourse Parser

TL;DR

This paper introduces a simple, neural, chart-based discourse parser that achieves state-of-the-art performance among global parsers without manual features, using an independence assumption for efficient decoding.

Contribution

It presents a novel global discourse parsing model that is simple, feature-free, and computationally efficient, outperforming previous global parsers.

Findings

Achieves best performance among global discourse parsers.

Comparable to state-of-the-art greedy parsers.

Uses only learned span representations without manual features.

Abstract

Discourse parsing is largely dominated by greedy parsers with manually-designed features, while global parsing is rare due to its computational expense. In this paper, we propose a simple chart-based neural discourse parser that does not require any manually-crafted features and is based on learned span representations only. To overcome the computational challenge, we propose an independence assumption between the label assigned to a node in the tree and the splitting point that separates its children, which results in tractable decoding. We empirically demonstrate that our model achieves the best performance among global parsers, and comparable performance to state-of-art greedy parsers, using only learned span representations.