Datasets:

HiTZ
/

wnli-eu

	---
	annotations_creators:
	- expert-generated
	language_creators:
	- expert-generated
	language:
	- eu
	license:
	- cc-by-4.0
	multilinguality:
	- monolingual
	size_categories:
	- n<1K
	source_datasets:
	- extended\|glue
	task_categories:
	- text-classification
	task_ids:
	- natural-language-inference
	pretty_name: WNLI-eu
	config_names:
	- wnli
	language_details: eu-ES
	configs:
	- config_name: wnli
	data_files:
	- split: validation
	path: wnli_validation.jsonl
	- split: test
	path: wnli_test.jsonl
	default: true
	dataset_info:
	config_name: wnli
	features:
	- name: sentence1
	dtype: string
	- name: sentence2
	dtype: string
	- name: label
	dtype:
	class_label:
	names:
	'0': not_entailment
	'1': entailment
	- name: idx
	dtype: int32
	splits:
	- name: validation
	num_bytes: 12960
	num_examples: 71
	- name: test
	num_bytes: 41450
	num_examples: 146
	download_size: 60587
	dataset_size: 54410
	---

	# Dataset Card for WNLI-eu

	- Point of Contact: [[email protected]](mailto:[email protected])

	## Dataset Description

	### Dataset Summary

	WNLI-eu is the professional translation to Basque of the WNLI dataset.
	WNLI is part of the GLUE benchmark for English ([Wang et al., 2018](https://aclanthology.org/W18-5446.pdf))
	and is based on the Winograd Schema Challenge (WSC) dataset ([Levesque et al., 2011](https://cdn.aaai.org/ocs/4492/4492-21843-1-PB.pdf)):
	> A Winograd schema is a pair of sentences differing in only one or two words and containing an ambiguity that is resolved in opposite ways in the two sentences and that requires the use of world knowledge and reasoning for its resolution.

	Specifically, WNLI instances consist of Winograd schemas converted into NLI sentence-pair classification problems.

	### Languages

	* eu-ES

	## Dataset Structure

	### Data Instances

	WNLI-eu examples look like this:

	```json
	{
	"sentence1": "Isurbidea ilearekin buxatuta dago. Garbitu egin behar da.",
	"sentence2": "Ilea garbitu behar da.",
	"label": 0,
	"idx": 0
	}
	```
	### Data Fields

	* `sentence1` (str): the premise.
	* `sentence2` (str): second sentence.
	* `label` (int): `1` or `0`, whether `sentence1` entails `sentence2` or not, respectively.
	* `idx` (int): index of the example.

	### Data Splits

	\| name \| validation \| test \|
	\|---------\|-----------:\|-----:\|
	\| default \| 71 \| 146 \|

	## Dataset Creation

	This dataset is a professional translation of the English WNLI dataset into Basque, commissioned by HiTZ (UPV/EHU) within the ILENIA project.
	For more information on how WNLI and the Winograd Schema Challenge dataset were created, please refer to their articles (see above).

	## Additional Information

	This work is funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the [project ILENIA](https://proyectoilenia.es/) with reference 2022/TL22/00215335.

	### Licensing Information

	[Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/)

	### Citation Information

	```
	@inproceedings{baucells-etal-2025-iberobench,
	title = "{I}bero{B}ench: A Benchmark for {LLM} Evaluation in {I}berian Languages",
	author = "Baucells, Irene and
	Aula-Blasco, Javier and
	de-Dios-Flores, Iria and
	Paniagua Su{\'a}rez, Silvia and
	Perez, Naiara and
	Salles, Anna and
	Sotelo Docio, Susana and
	Falc{\~a}o, J{\'u}lia and
	Saiz, Jose Javier and
	Sepulveda Torres, Robiert and
	Barnes, Jeremy and
	Gamallo, Pablo and
	Gonzalez-Agirre, Aitor and
	Rigau, German and
	Villegas, Marta",
	editor = "Rambow, Owen and
	Wanner, Leo and
	Apidianaki, Marianna and
	Al-Khalifa, Hend and
	Eugenio, Barbara Di and
	Schockaert, Steven",
	booktitle = "Proceedings of the 31st International Conference on Computational Linguistics",
	month = jan,
	year = "2025",
	address = "Abu Dhabi, UAE",
	publisher = "Association for Computational Linguistics",
	url = "https://aclanthology.org/2025.coling-main.699/",
	pages = "10491--10519",
	}
	```