XNLIvar
收藏资源简介:
XNLIvar数据集是首个公开的手工整理的NLI数据集,用于评估自然语言处理技术对巴斯克语和西班牙语地区语言变体的理解能力。该数据集由HiTZ中心与巴塞罗那大学合作开发,包含了巴斯克语和西班牙语的多种地区语言变体,旨在帮助NLP系统更好地处理不同地区的语言变体。数据集由12位巴斯克语母语者和6位西班牙语母语者参与创建,涵盖了巴斯克语和西班牙语的多个地区变体,包括西部、中部和纳瓦拉地区。数据集共包含1550条数据,其中巴斯克语变体数据894条,西班牙语变体数据666条。该数据集可用于训练和评估自然语言理解模型,以解决跨地区语言变体带来的挑战。
XNLIvar is the first publicly available manually curated Natural Language Inference (NLI) dataset developed to evaluate the capability of natural language processing (NLP) technologies in understanding regional language varieties of Basque and Spanish. Developed in collaboration between the HiTZ Center and the University of Barcelona, this dataset covers multiple regional language varieties of Basque and Spanish, aiming to help NLP systems better handle cross-regional language variations. It was created with the participation of 12 native Basque speakers and 6 native Spanish speakers, encompassing regional varieties from the Western, Central, and Navarre regions of both languages. In total, the dataset contains 1,550 instances, including 894 instances of Basque language varieties and 666 instances of Spanish language varieties. This dataset can be utilized for training and evaluating natural language understanding models to address the challenges brought by cross-regional language variations.
XNLIvar数据集概述
基本信息
- 许可证: CC-BY-NC-4.0
- 语言: 西班牙语(es)、巴斯克语(eu)
- 标签: Variation
- 数据集名称: xnli_var
数据集结构
配置(configs)
-
eu配置
- original: eu/xnli-eu-original.tsv
- native: eu/xnli-eu-native.tsv
- variation: eu/xnli-eu-var.tsv
-
es配置
- original: es/xnli-es-original.tsv
- native: es/xnli-eu2es-native.tsv
- variation: es/xnli-es-var.tsv
-
other_ablation配置
- Basque_no_repetition: ablation-eu/xnli-native-var-eu-NO-REPETITION.tsv
- Basque_less_western: ablation-eu/xnli-native-var-eu-less-biz.tsv
- Basque_less_central: ablation-eu/xnli-native-var-eu-less-gip.tsv
- Spanish_no_repetition: ablation-es/xnli-native-var-es-no-rep.tsv
-
translations配置
- translations_Basque_native: translations/xnli-eu2en-native.tsv
- translations_Basque_variation: translations/xnli-eu2en-var.tsv
- translation_Spanish_native: translations/xnli-es2en-native.tsv
- translation_Spanish_variation: translations/xnli-es2en-var.tsv
数据集描述
- 来源: 衍生自XNLI数据集,采用相同许可证
- 特点: 手工整理的巴斯克语和西班牙语变体包容性NLI数据集
- 用途: 评估语言技术对巴斯克语和西班牙语变体的理解能力
文件夹结构
- eu: 包含原始XNLI测试数据、原生和变体包容性数据集
- es: 包含原始XNLI测试数据、巴斯克原生数据翻译成西班牙语版本及变体包容性数据
- Translations: 提供原生和变体包容性数据集的自动英语翻译
- ablation-eu: 巴斯克语消融实验数据集
- ablation-es: 西班牙语消融实验数据集
相关资源
- GitHub仓库: https://github.com/hitz-zentroa/XNLIvar
- 论文链接: https://arxiv.org/abs/2506.15239
引用格式
bibtex @inproceedings{bengoetxea-et-al-2025,, title = "Lost in Variation? Evaluating NLI Performance in Basque and Spanish Geographical Variants", author = "Bengoetxea, Jaione and Gonzalez-Dios, Itziar and Agerri, Rodrigo", year = "2025", url = "https://arxiv.org/abs/2506.15239" }

- 1Lost in Variation? Evaluating NLI Performance in Basque and Spanish Geographical VariantsHiTZ Center - Ixa, University of the Basque Country UPV/EHU · 2025年



