XQuAD (Cross-lingual Question Answering Dataset): 1,190 SQuAD Questions Professionally Translated into 11 Languages (Google DeepMind)
- Provider
- DeepMind (Artetxe, Ruder and Yogatama)
- License
- CC-BY-SA-4.0 (source)
- Size
- 14 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/google/xquad)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Natural Language Processing
- Source
- https://arxiv.org/abs/1910.11856
About this dataset
XQuAD (Cross-lingual Question Answering Dataset) evaluates cross-lingual extractive question answering. It takes 240 paragraphs and 1,190 question-answer pairs from the SQuAD v1.1 development set and adds professional human translations into Spanish, German, Greek, Russian, Turkish, Arabic, Vietnamese, Thai, Chinese and Hindi, so the set is fully parallel across 11 languages; a Romanian translation was added later. Each example has a context paragraph, a question and the answer span with its character offset, in SQuAD format. It is the standard zero-shot transfer test for multilingual language models (mBERT, XLM-R, mT5): train on English SQuAD, evaluate in each target language. The repository has one folder per language (xquad.ar, xquad.de, xquad.zh and so on), each with a validation Parquet file. Load with the Hugging Face datasets library (load_dataset('google/xquad', 'xquad.de')). Distributed by DeepMind under CC BY-SA 4.0; cite Artetxe, Ruder and Yogatama, On the Cross-lingual Transferability of Monolingual Representations, ACL 2020.
How to load this dataset in Python
from datasets import load_dataset
ds = load_dataset("google/xquad")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: xquad.ar/validation-00000-of-00001.parquet
Analyze this dataset without downloading it
On Scigantic, XQuAD (Cross-lingual Question Answering Dataset): 1,190 SQuAD Questions Professionally Translated into 11 Languages (Google DeepMind) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Natural Language Processing datasets
- Md Gender Bias
- S2AND: Author Name Disambiguation Benchmark, Eight Bibliographic Datasets plus Semantic Scholar Production Models (Allen Institute for AI)
- IndicGenBench XQuAD-IN: Extractive Question Answering Translated into Indic Languages (Google Research)
- BFCL Hi
- BLiMP: Benchmark of Linguistic Minimal Pairs for English Grammar (67 Phenomena, NYU)
- PASS: Perturb-and-Select Summarizer for Product Reviews
- Nemotron Safety Guard Dataset V 3
- Asset