README.md

metadata

license: apache-2.0

annotations_creators:

• found language_creators:
• expert-generated language:
• eng
• amh
• ibo
• fra
• sna
• lin
• wol
• ewe
• lug
• xho
• kin
• twi
• zul
• orm
• yor
• hau
• sot
• swa multilinguality:
• multilingual pretty_name: afrixnli language_details: eng, amh, ibo, fra, sna, lin, wol, ewe, lug, xho, kin, twi, zul, orm, yor, hau, sot, swa size_categories: 200<n<1K source_datasets:
• xnli tags:
• afrixnli
• afri-xnli
• africanxnli task_categories:
• text-classification task_ids:
• text-classification

---

Dataset Card for afrixnli

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Point of Contact: d.adelani@ucl.ac.uk

Dataset Summary

AFRIXNLI is an evaluation dataset comprising translations of a subset of the XNLI dataset into 16 African languages. It includes both validation and test sets across all 18 languages, maintaining the English and French subsets from the original XNLI dataset.

Languages

There are 18 languages available :

Dataset Structure

Data Instances

The examples look like this for English:

from datasets import load_dataset
data = load_dataset('masakhane/afrixnli', 'eng') 
# Please, specify the language code
# A data point example is below:
{
'premise': 'The doors were locked when we went in.',
'hypothesis': 'All of the doors were open.',
'label': 'contradiction'
}

Data Fields

• premise: a multilingual string variable,
• hypothesis: a multilingual string variable,
• label: a classification label, with possible values including entailment (0), neutral (1), contradiction (2).

Data Splits

All languages has two splits, dev and test a subset of the original dev and test splits of the XNLI dataset.

The splits have the following sizes :

Language	validation	test
English	450	600