Identifying Source-Language Dialects in Translation

In this paper, we aim to explore the degree to which translated texts preserve linguistic features of dialectal varieties. We release a dataset of augmented annotations to the Proceedings of the European Parliament that cover dialectal speaker information, and we analyze different classes of written English covering native varieties from the British Isles. Our analyses aim to discuss the discriminatory features between the different classes and to reveal words whose usage differs between varieties of the same language. We perform classification experiments and show that automatically distinguishing between the dialectal varieties is possible with high accuracy, even after translation, and propose a new explainability method based on embedding alignments in order to reveal specific differences between dialects at the level of the vocabulary.