A Simple Global Neural Discourse Parser
This addresses the problem of inefficient global parsing for NLP researchers, offering a more scalable approach, though it is incremental as it builds on existing parsing methods.
The paper tackles the computational expense of global discourse parsing by proposing a chart-based neural parser that uses learned span representations and an independence assumption for tractable decoding, achieving the best performance among global parsers and comparable results to state-of-the-art greedy parsers.
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.