Machine Translation Robustness to Natural Asemantic Variation

TL;DR

This paper introduces Natural Asemantic Variation (NAV) as a new challenge for machine translation, demonstrating that current models struggle with subtle, meaning-preserving language nuances and proposing fine-tuning strategies to improve robustness across languages.

Contribution

The paper formalizes NAV as a new category of variation affecting MT robustness and explores methods to enhance model performance using human-generated and synthetic variations.

Findings

Existing MT models fail on NAV data

Fine-tuning with human variations improves NAV robustness

Synthetic perturbations partially replicate organic NAV benefits

Abstract

Current Machine Translation (MT) models still struggle with more challenging input, such as noisy data and tail-end words and phrases. Several works have addressed this robustness issue by identifying specific categories of noise and variation then tuning models to perform better on them. An important yet under-studied category involves minor variations in nuance (non-typos) that preserve meaning w.r.t. the target language. We introduce and formalize this category as Natural Asemantic Variation (NAV) and investigate it in the context of MT robustness. We find that existing MT models fail when presented with NAV data, but we demonstrate strategies to improve performance on NAV by fine-tuning them with human-generated variations. We also show that NAV robustness can be transferred across languages and find that synthetic perturbations can achieve some but not all of the benefits of organic NAV data.