Gender issues in machine translation

Abstract

Recent years have seen significant improvements in machine translation thanks to the enhanced capacity of computers and advanced technologies in the field of Natural Language Processing (NLP). Our contribution focuses on one of the most frequent problems in the state-of-the-art MT systems: translation of gender. It represents a recurrent source of mistranslation: incorrect gender attribution to proforms (personal pronouns, relative pronouns, etc.), the reproduction of gender stereotypes and the overuse of masculine pronouns are among the most frequent problems. In this chapter we briefly outline the different approaches to MT, we address historical perspectives and recent developments concerning gender issues in MT. The final part of the chapter is devoted to a brief discussion of outstanding issues in this field, namely the social and ethical impact of an unbalanced representation of gender in MT output and finally the need of i) fine-grained evaluation metrics which help focus on specific critical areas and ii) benchmark datasets specifically devoted to gender handling.