The Impact of Edge Displacement Vaserstein Distance on UD Parsing Performance

TL;DR

This paper introduces a new measurement based on edge displacement Wasserstein distance to evaluate how distributional differences between training and test data affect NLP parsing performance, providing insights into system bounds.

Contribution

It presents a novel statistical measurement of edge displacement distribution differences and demonstrates its correlation with parsing performance, offering a new tool for NLP analysis.

Findings

Statistical correlation between edge displacement Wasserstein distance and parsing accuracy.

A sampling technique to estimate lower and upper bounds of parsing system performance.

Methodology can serve as a reference for future correlation-based NLP research.

Abstract

We contribute to the discussion on parsing performance in NLP by introducing a measurement that evaluates the differences between the distributions of edge displacement (the directed distance of edges) seen in training and test data. We hypothesize that this measurement will be related to differences observed in parsing performance across treebanks. We motivate this by building upon previous work and then attempt to falsify this hypothesis by using a number of statistical methods. We establish that there is a statistical correlation between this measurement and parsing performance even when controlling for potential covariants. We then use this to establish a sampling technique that gives us an adversarial and complementary split. This gives an idea of the lower and upper bounds of parsing systems for a given treebank in lieu of freshly sampled data. In a broader sense, the methodology presented here can act as a reference for future correlation-based exploratory work in NLP.