thoshiths/miracl-multilingual-4M

MIRACL Multilingual Wikipedia Corpus — 4M Passages

A curated 4-million-passage multilingual Wikipedia corpus derived from the MIRACL dataset, sampled for use in Information Retrieval research and assignments.

Languages

Language	Code	Passages	Source corpus size
English	en	1,000,000	~32.9M
Spanish	es	1,000,000	~10.4M
French	fr	1,000,000	~14.6M
German	de	1,000,000	~15.9M

Total: 4,000,000 passages

Dataset Structure

{
    "doc_id":       "en_0000042",     # unique passage ID
    "title":        "Alan Turing",    # Wikipedia article title
    "text":         "Alan Mathison…", # passage text
    "language":     "en",             # ISO 639-1 language code
    "url":          "https://en.wikipedia.org/?curid=1208",
    "miracl_docid": "1208#0",         # original MIRACL docid (article#passage)
}

Usage

from datasets import load_dataset

# Load full dataset
ds = load_dataset("thoshiths/miracl-multilingual-4M", split="train")

# Filter by language
en_docs = ds.filter(lambda x: x["language"] == "en")

Source Data

Passages are sampled from the MIRACL Wikipedia corpora (Zhang et al., 2023) using reservoir sampling to ensure representative coverage.

Original data is derived from Wikipedia dumps and licensed under CC BY-SA 4.0.

Citation

If you use this dataset, please cite the original MIRACL paper:

@article{zhang2023miracl,
  title   = {MIRACL: A Multilingual Retrieval Dataset Covering 18 Diverse Languages},
  author  = {Xinyu Zhang and Nandan Thakur and Odunayo Ogundepo and Ehsan Kamalloo
             and David Alfonso-Hermelo and Xiaoguang Li and Qun Liu
             and Mehdi Rezagholizadeh and Jimmy Lin},
  journal = {Transactions of the Association for Computational Linguistics},
  year    = {2023},
}