Datasets:
metadata
license: apache-2.0
mutilinguality:
- multilingual
task_categories:
- text-generation
task_ids:
- language-modeling
language:
- afr
- amh
- arz
- eng
- fra
- hau
- ibo
- kin
- mlg
- nya
- orm
- por
- sna
- som
- sot
- swa
- tir
- xho
- yor
- zul
viewer: true
dataset_info:
- config_name: afr
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 4549624636
num_examples: 1042812
- name: validation
num_bytes: 504320368
num_examples: 115868
download_size: 5124049817
dataset_size: 5053945004
- config_name: amh
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 1115662532
num_examples: 135863
- name: validation
num_bytes: 123858179
num_examples: 15095
download_size: 1248728162
dataset_size: 1239520711
- config_name: arz
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 621073489
num_examples: 1455662
- name: validation
num_bytes: 69342976
num_examples: 161740
download_size: 753246622
dataset_size: 690416465
- config_name: eng
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 3795223480
num_examples: 1378555
- name: validation
num_bytes: 423622310
num_examples: 153172
download_size: 4279723559
dataset_size: 4218845790
- config_name: fra
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 3340740638
num_examples: 1443177
- name: validation
num_bytes: 368983958
num_examples: 160352
download_size: 3796280757
dataset_size: 3709724596
- config_name: hau
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 909342448
num_examples: 359881
- name: validation
num_bytes: 101151882
num_examples: 39986
download_size: 1027800797
dataset_size: 1010494330
- config_name: ibo
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 193493918
num_examples: 51386
- name: validation
num_bytes: 22265232
num_examples: 5709
download_size: 219266571
dataset_size: 215759150
- config_name: kin
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 208582172
num_examples: 97064
- name: validation
num_bytes: 10662209
num_examples: 5831
download_size: 222938591
dataset_size: 219244381
- config_name: mlg
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 561868602
num_examples: 216210
- name: validation
num_bytes: 62280728
num_examples: 24023
download_size: 635783521
dataset_size: 624149330
- config_name: nya
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 260737793
num_examples: 39647
- name: validation
num_bytes: 29199589
num_examples: 4405
download_size: 293880333
dataset_size: 289937382
- config_name: orm
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 51725718
num_examples: 20169
- name: validation
num_bytes: 5500617
num_examples: 2241
download_size: 58001407
dataset_size: 57226335
- config_name: por
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 2191644027
num_examples: 1089199
- name: validation
num_bytes: 245338209
num_examples: 121022
download_size: 2498665351
dataset_size: 2436982236
- config_name: sna
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 225393219
num_examples: 60986
- name: validation
num_bytes: 25595688
num_examples: 6776
download_size: 254964089
dataset_size: 250988907
- config_name: som
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 2165910731
num_examples: 976484
- name: validation
num_bytes: 241175779
num_examples: 108498
download_size: 2451878912
dataset_size: 2407086510
- config_name: sot
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 199386007
num_examples: 38361
- name: validation
num_bytes: 22324957
num_examples: 4262
download_size: 224556522
dataset_size: 221710964
- config_name: swa
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 3371589021
num_examples: 1036254
- name: validation
num_bytes: 373326029
num_examples: 115139
download_size: 3804265021
dataset_size: 3744915050
- config_name: tir
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 32026542
num_examples: 8240
- name: validation
num_bytes: 3589604
num_examples: 915
download_size: 35955368
dataset_size: 35616146
- config_name: xho
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 114450184
num_examples: 23892
- name: validation
num_bytes: 13051255
num_examples: 2654
download_size: 129410950
dataset_size: 127501439
- config_name: yor
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 192473693
num_examples: 73473
- name: validation
num_bytes: 21123764
num_examples: 8163
download_size: 217343993
dataset_size: 213597457
- config_name: zul
features:
- name: id
dtype: string
- name: headline
dtype: string
- name: content
dtype: string
- name: category
dtype: string
- name: url
dtype: string
splits:
- name: train
num_bytes: 279244495
num_examples: 65447
- name: validation
num_bytes: 30487397
num_examples: 7271
download_size: 314070508
dataset_size: 309731892
Dataset Summary
WURA is a document-level dataset covering 16 African Languages and 4 high-resource languages widely spoken in Africa (English, French, Arabic and Portuguese). This dataset was created by auditing mC4 and crawling additional verified news sources. It was first used to train AfriTeVa V2.
Dataset Structure
>>> from datasets import load_dataset
Although the document-level dataset is loaded by default, you may also optionally load a passage-level dataset as follows
>>> data = load_dataset("castorini/wura, "yor", level="passage", verification_mode="no_checks")
Note that we must pass verification_mode="no_checks to prevent HF from verifying checksums against the document-level checksum infos.
Citation
@inproceedings{oladipo-etal-2023-better,
title = "Better Quality Pre-training Data and T5 Models for {A}frican Languages",
author = "Oladipo, Akintunde and
Adeyemi, Mofetoluwa and
Ahia, Orevaoghene and
Owodunni, Abraham and
Ogundepo, Odunayo and
Adelani, David and
Lin, Jimmy",
editor = "Bouamor, Houda and
Pino, Juan and
Bali, Kalika",
booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.emnlp-main.11",
pages = "158--168",
abstract = "In this study, we highlight the importance of enhancing the quality of pretraining data in multilingual language models. Existing web crawls have demonstrated quality issues, particularly in the context of low-resource languages. Consequently, we introduce a new multilingual pretraining corpus for 16 African languages, designed by carefully auditing existing pretraining corpora to understand and rectify prevalent quality issues. To compile this dataset, we undertake a rigorous examination of current data sources for thirteen languages within one of the most extensive multilingual web crawls, mC4, and extract cleaner data through meticulous auditing and improved web crawling strategies. Subsequently, we pretrain a new T5-based model on this dataset and evaluate its performance on multiple downstream tasks. Our model demonstrates better downstream effectiveness over existing pretrained models across four NLP tasks, underscoring the critical role data quality plays in pretraining language models in low-resource scenarios. Specifically, on cross-lingual QA evaluation, our new model is more than twice as effective as multilingual T5. All code, data and models are publicly available at https://github.com/castorini/AfriTeVa-keji.",
}