Improving Coverage and Runtime Complexity for Exact Inference in Non-Projective Transition-Based Dependency Parsers

TL;DR

This paper introduces new non-projective transition-based dependency parsers that achieve polynomial-time exact inference, improving coverage and reducing runtime complexity, thus advancing the theoretical understanding and practical efficiency of dependency parsing.

Contribution

It generalizes existing parsers to a family with better coverage and lower time complexity, including a novel variant with $O(n^6)$ runtime.

Findings

New parsers with improved coverage over previous models

A variant achieving $O(n^6)$ time complexity for exact inference

Theoretical framework inspiring future parser design

Abstract

We generalize Cohen, G\'omez-Rodr\'iguez, and Satta's (2011) parser to a family of non-projective transition-based dependency parsers allowing polynomial-time exact inference. This includes novel parsers with better coverage than Cohen et al. (2011), and even a variant that reduces time complexity to $O(n^6)$, improving over the known bounds in exact inference for non-projective transition-based parsing. We hope that this piece of theoretical work inspires design of novel transition systems with better coverage and better run-time guarantees. Code available at https://github.com/tzshi/nonproj-dp-variants-naacl2018