license: odc-by
language:
- en
tags:
- gender
- ambiguity
pretty_name: 'GAND: Gender Ambiguous Natural Data'
size_categories:
- 1K<n<10K
task_categories:
- translation
Dataset Description
GAND (Gender-Ambiguous Natural Data) is a benchmarking resource for evaluating gender (bias) in machine translation or downstream NLP tasks. The data stems purely from natural data resources (OpenSubtitles from the OPUS project and C4). The data has been meticulously (automatically + manually) filtered to ensure complete gender ambiguity with respect to a specific referent. More information on the compilation of GAND can be found on GitHub.
Usage
from datasets import load_dataset
ds = load_dataset("jhacken/GAND")
Train, dev, test split
- Train set: 4037 rows
- Dev set: 505 rows
- Test set: 505 rows
Dataset Structure
| referent | EN_source_sentence | referent_embedding | sentence_source |
|---|---|---|---|
| assistant | No one' s permitted to enter the library... other than myself and my assistant. | female_embedding_list | OpenSubtitles |
| specialist | As a social media specialist with a million things on your plate, you might not have been aware that citrus was all the rage atm. | LLM_neutral_list | C4 |
Cite this dataset
@dataset{hackenbuchner_2026_20324375, author = {Hackenbuchner, Janiça and Degraeuwe, Jasper and Tezcan, Arda and Daems, Joke}, title = {GAND Dataset: Gender-Ambiguous Natural Data}, month = may, year = 2026, publisher = {Zenodo}, version = {v1.0.0}, doi = {10.5281/zenodo.20324375}, url = {https://doi.org/10.5281/zenodo.20324375}, }
Acknowledgements
GAND was developed as part of a strategic basic PhD research (1SH5V24N) fully funded by The Research Foundation – Flanders (FWO) for the time span of four years, from 01.11.2023 until 31.10.2027, and hosted within the Language and Translation Technology Team (LT3) at Ghent University. The computational resources (Stevin Supercomputer Infrastructure) and services used in this work were provided by the VSC (Flemish Supercomputer Center), funded by Ghent University, FWO and the Flemish Government - department EWI.