Linear-Time Constituency Parsing with RNNs and Dynamic Programming

TL;DR

This paper introduces a linear-time constituency parser using RNNs and dynamic programming, significantly improving speed over traditional cubic-time parsers while maintaining high accuracy, especially for long sentences and discourse parsing.

Contribution

The paper presents a novel linear-time constituency parser with RNNs and dynamic programming, reducing parsing complexity from cubic to near-linear time with beam search and cube pruning.

Findings

Runs in $O(n b \,\log b)$ time, faster than traditional methods.

Achieves highest F1 accuracy on Penn Treebank among single end-to-end models.

Substantially faster for long sentences and discourse parsing.

Abstract

Recently, span-based constituency parsing has achieved competitive accuracies with extremely simple models by using bidirectional RNNs to model "spans". However, the minimal span parser of Stern et al (2017a) which holds the current state of the art accuracy is a chart parser running in cubic time, $O(n^3)$, which is too slow for longer sentences and for applications beyond sentence boundaries such as end-to-end discourse parsing and joint sentence boundary detection and parsing. We propose a linear-time constituency parser with RNNs and dynamic programming using graph-structured stack and beam search, which runs in time $O(n b^2)$ where $b$ is the beam size. We further speed this up to $O(n b\log b)$ by integrating cube pruning. Compared with chart parsing baselines, this linear-time parser is substantially faster for long sentences on the Penn Treebank and orders of magnitude faster for discourse parsing, and achieves the highest F1 accuracy on the Penn Treebank among single model end-to-end systems.