Skip to content

Dataset · MoritzLaurer

Multilingual NLI corpus across 26 languages (2.73M pairs)

MoritzLaurer/multilingual-NLI-26lang-2mil7

Massive set of 2.73 million text pairs in 26 languages for training logical inference and zero-shot classification models.

What's inside

The dataset contains 2,730,000 text pairs for the Natural Language Inference (NLI) task across 26 languages (including Chinese, Japanese, Arabic, Korean, German, and others). Created using open machine learning translation models based on well-known English corpora MultiNLI, Fever-NLI, ANLI, LingNLI, and WANLI.

What you can do with it

  • Training multilingual models for Natural Language Inference (NLI)
  • Training text classifiers in zero-shot mode for 26 languages
  • Evaluating cross-lingual natural language understanding quality
Tasks
Text classification
Languages
Multilingual, Chinese, Japanese, Arabic, Korean, German, French, Spanish and 4 more
Size
1M–10M rows
Downloads this month
1.7K
Likes
52

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("MoritzLaurer/multilingual-NLI-26lang-2mil7")