Dependency Parsing as Head Selection

TL;DR

This paper introduces DeNSe, a neural head selection model for dependency parsing that predicts heads independently, producing mostly valid trees without structural constraints during training, and achieves competitive results across multiple languages.

Contribution

The paper proposes a novel head selection approach for dependency parsing that simplifies training and maintains high accuracy without enforcing tree constraints during learning.

Findings

DeNSe produces valid dependency trees for most sentences.

The model achieves state-of-the-art performance on multiple languages.

Non-tree outputs can be corrected with a maximum spanning tree algorithm.

Abstract

Conventional graph-based dependency parsers guarantee a tree structure both during training and inference. Instead, we formalize dependency parsing as the problem of independently selecting the head of each word in a sentence. Our model which we call \textsc{DeNSe} (as shorthand for {\bf De}pendency {\bf N}eural {\bf Se}lection) produces a distribution over possible heads for each word using features obtained from a bidirectional recurrent neural network. Without enforcing structural constraints during training, \textsc{DeNSe} generates (at inference time) trees for the overwhelming majority of sentences, while non-tree outputs can be adjusted with a maximum spanning tree algorithm. We evaluate \textsc{DeNSe} on four languages (English, Chinese, Czech, and German) with varying degrees of non-projectivity. Despite the simplicity of the approach, our parsers are on par with the state of the art.