Efficient EUD Parsing

TL;DR

This paper introduces an efficient dependency parsing system combining neural and rule-based methods, achieving competitive accuracy with reduced training and inference costs for the EUD Shared Task at IWPT 2020.

Contribution

It presents a novel combination of distilled neural parsers and rule-based projection techniques for efficient EUD parsing.

Findings

Achieved an average ELAS of 74.04

Ranked 4th overall in the shared task

Demonstrated effective efficiency improvements

Abstract

We present the system submission from the FASTPARSE team for the EUD Shared Task at IWPT 2020. We engaged with the task by focusing on efficiency. For this we considered training costs and inference efficiency. Our models are a combination of distilled neural dependency parsers and a rule-based system that projects UD trees into EUD graphs. We obtained an average ELAS of 74.04 for our official submission, ranking 4th overall.