FBK-MT/GNR-it

---
license: cc-by-4.0
configs:
- config_name: default
data_files:
- split: clean
path: data/clean-*
- split: full
path: data/full-*
dataset_info:
features:
- name: id
dtype: int64
- name: gendered
dtype: string
- name: neutral
dtype: string
splits:
- name: clean
num_bytes: 25496599
num_examples: 81389
- name: full
num_bytes: 50772270
num_examples: 162778
download_size: 28044975
dataset_size: 76268869
task_categories:
- text-classification
- text-generation
language:
- it
tags:
- fairness
- rewriting
- gender-inclusive
- gender-neutral
size_categories:
- 100K<n<1M
---

# GNR-it Dataset

## Table of Contents
1. [Overview](#overview)
2. [Usage](#usage)
3. [License](#license)
4. [Citation](#citation)

## Overview

The **GNR-it** dataset contains pairs of gendered and gender-neutral Italian sentences.

We release this dataset to ensure reproducibility of the experiments in the paper [Gender-Neutral Rewriting in Italian: Models, Approaches, and Trade-offs](https://arxiv.org/abs/2509.13480), accepted at CLiC-it 2025.

The dataset is derived from the data originally created to train the gender-neutrality classifier [GeNTE-evaluator](https://huggingface.co/FBK-MT/GeNTE-evaluator).
The creation and curation of the original dataset is described in the paper [Hi Guys or Hi Folks? Benchmarking Gender-Neutral Machine Translation with the GeNTE Corpus (Piergentili et al., 2023)](https://aclanthology.org/2023.emnlp-main.873/).

Entries in this dataset include the following columns:
* **id**: a progressive identifier
* **gendered**: the gendered sentence
* **neutral**: the gender-neutral sentence

To facilitate reproducibility of our paper’s experiments, we release both splits:
- **full**: the complete set of 162,778 pairs
- **clean**: a subset of 81,389 pairs selected based on their BERTScore

These two splits dataset were used to fine-tune the following models:
- [Qwen3-8B-GNR-it-clean](https://huggingface.co/FBK-MT/Qwen3-8B-GNR-it-clean)
- [Qwen3-8B-GNR-it-full](https://huggingface.co/FBK-MT/Qwen3-8B-GNR-it-full)
- [Qwen3-14B-GNR-it-clean](https://huggingface.co/FBK-MT/Qwen3-14B-GNR-it-clean)
- [Qwen3-14B-GNR-it-full](https://huggingface.co/FBK-MT/Qwen3-14B-GNR-it-full)

## Usage

```python
from datasets import load_dataset

# Full set
full_data = load_dataset("FBK-MT/GNR-it", split="full")

# Clean set
clean_data = load_dataset("FBK-MT/GNR-it", split="clean")
```

## License

We release this dataset under the Creative Commons Attribution 4.0 International license (CC BY 4.0).

## Citation

If you this dataset in your work, please cite:

```
@misc{piergentili2025genderneutralrewritingitalianmodels,
title={Gender-Neutral Rewriting in Italian: Models, Approaches, and Trade-offs},
author={Andrea Piergentili and Beatrice Savoldi and Matteo Negri and Luisa Bentivogli},
year={2025},
eprint={2509.13480},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2509.13480},
}
```

## Contributions

Thanks to [@apiergentili](https://huggingface.co/apiergentili) for adding this dataset.