Neural Generative Rhetorical Structure Parsing

TL;DR

This paper introduces the first generative RST parser using an RNN grammar, improving accuracy over previous models and addressing biases in beam search algorithms for document-level rhetorical structure parsing.

Contribution

It presents a novel generative RST parsing model with a new beam search algorithm that reduces bias and enhances parsing performance.

Findings

Outperforms previous algorithms with 6.8 and 2.9 F1 improvements

Outperforms discriminative models with 2.6 F1 points

Achieves state-of-the-art performance without extra training data

Abstract

Rhetorical structure trees have been shown to be useful for several document-level tasks including summarization and document classification. Previous approaches to RST parsing have used discriminative models; however, these are less sample efficient than generative models, and RST parsing datasets are typically small. In this paper, we present the first generative model for RST parsing. Our model is a document-level RNN grammar (RNNG) with a bottom-up traversal order. We show that, for our parser's traversal order, previous beam search algorithms for RNNGs have a left-branching bias which is ill-suited for RST parsing. We develop a novel beam search algorithm that keeps track of both structure- and word-generating actions without exhibiting this branching bias and results in absolute improvements of 6.8 and 2.9 on unlabelled and labelled F1 over previous algorithms. Overall, our generative model outperforms a discriminative model with the same features by 2.6 F1 points and achieves performance comparable to the state-of-the-art, outperforming all published parsers from a recent replication study that do not use additional training data.