Synthetic and Natural Noise Both Break Neural Machine Translation

Yonatan Belinkov, Yonatan Bisk

Abstract: Character-based neural machine translation (NMT) models alleviate out-of-vocabulary issues, learn morphology, and move us closer to completely end-to-end translation systems. Unfortunately, they are also very brittle and easily falter when presented with noisy data. In this paper, we confront NMT models with synthetic and natural sources of noise. We find that state-of-the-art models fail to translate even moderately noisy texts that humans have no trouble comprehending. We explore two approaches to increase model robustness: structure-invariant word representations and robust training on noisy texts. We find that a model based on a character convolutional neural network is able to simultaneously learn representations robust to multiple kinds of noise.

TL;DR: CharNMT is brittle

Keywords: neural machine translation, characters, noise, adversarial examples, robust training

Code: [ ybisk/charNMT-noise](https://github.com/ybisk/charNMT-noise) + [ 2 community implementations](https://paperswithcode.com/paper/?openreview=BJ8vJebC-)

Community Implementations: [ 2 code implementations](https://www.catalyzex.com/paper/arxiv:1711.02173/code)