Do Not Change Me: On Transferring Entities Without Modification in Neural Machine Translation -- a Multilingual Perspective

TL;DR

This paper evaluates how well multilingual NMT models preserve specific entities like URLs and emails during translation, identifies challenges such as emojis, and introduces a new dataset for assessing entity transfer quality.

Contribution

It provides a comprehensive analysis of entity preservation in NMT models and introduces a synthetic dataset for evaluating entity transfer across multiple languages and categories.

Findings

Models vary in accuracy for entity preservation.

Emojis significantly challenge current models.

The new dataset enables systematic evaluation of entity transfer.

Abstract

Current machine translation models provide us with high-quality outputs in most scenarios. However, they still face some specific problems, such as detecting which entities should not be changed during translation. In this paper, we explore the abilities of popular NMT models, including models from the OPUS project, Google Translate, MADLAD, and EuroLLM, to preserve entities such as URL addresses, IBAN numbers, or emails when producing translations between four languages: English, German, Polish, and Ukrainian. We investigate the quality of popular NMT models in terms of accuracy, discuss errors made by the models, and examine the reasons for errors. Our analysis highlights specific categories, such as emojis, that pose significant challenges for many models considered. In addition to the analysis, we propose a new multilingual synthetic dataset of 36,000 sentences that can help assess the quality of entity transfer across nine categories and four aforementioned languages.