Evaluating Machine Translation Datasets for Low-Web Data Languages: A Gendered Lens
This work addresses the problem of harmful biases in datasets for low-resourced languages, which can lead to poor-performing and biased NLP technologies, and is incremental in highlighting specific dataset flaws.
The paper investigated the quality of machine translation datasets for three low-resourced languages (Afan Oromo, Amharic, Tigrinya), focusing on gender representation, and found significant male gender skews and harmful depictions against women, with these issues more prominent in languages with larger datasets.
As low-resourced languages are increasingly incorporated into NLP research, there is an emphasis on collecting large-scale datasets. But in prioritizing quantity over quality, we risk 1) building language technologies that perform poorly for these languages and 2) producing harmful content that perpetuates societal biases. In this paper, we investigate the quality of Machine Translation (MT) datasets for three low-resourced languages--Afan Oromo, Amharic, and Tigrinya, with a focus on the gender representation in the datasets. Our findings demonstrate that while training data has a large representation of political and religious domain text, benchmark datasets are focused on news, health, and sports. We also found a large skew towards the male gender--in names of persons, the grammatical gender of verbs, and in stereotypical depictions in the datasets. Further, we found harmful and toxic depictions against women, which were more prominent for the language with the largest amount of data, underscoring that quantity does not guarantee quality. We hope that our work inspires further inquiry into the datasets collected for low-resourced languages and prompts early mitigation of harmful content. WARNING: This paper contains discussion of NSFW content that some may find disturbing.