Scaling Graph-Based Dependency Parsing with Arc Vectorization and Attention-Based Refinement

TL;DR

This paper introduces a unified graph-based dependency parser that uses arc vectorization and attention mechanisms to improve accuracy and scalability, outperforming existing models on PTB and UD datasets.

Contribution

The novel architecture unifies arc scoring and labeling, overcomes scalability issues, and incorporates transformer layers for higher-order dependency modeling.

Findings

Outperforms state-of-the-art parsers in accuracy

Enhances scalability and efficiency

Effectively models higher-order dependencies

Abstract

We propose a novel architecture for graph-based dependency parsing that explicitly constructs vectors, from which both arcs and labels are scored. Our method addresses key limitations of the standard two-pipeline approach by unifying arc scoring and labeling into a single network, reducing scalability issues caused by the information bottleneck and lack of parameter sharing. Additionally, our architecture overcomes limited arc interactions with transformer layers to efficiently simulate higher-order dependencies. Experiments on PTB and UD show that our model outperforms state-of-the-art parsers in both accuracy and efficiency.