Hugging Face's logo

Datasets:
GEM
/
xlsum

Tasks:
Summarization
Languages:
Amharic
Arabic
Azerbaijani
ArXiv:
License:
Dataset card Files Community
5
xlsum / README.md
Tahmid's picture
Tahmid
Update README.md
dbccc03
preview code
|
raw
history blame
17.3 kB
metadata
task_categories:
  - conditional-text-generation
task_ids:
  - summarization
languages:
  - am
  - ar
  - az
  - bn
  - my
  - zh
  - en
  - fr
  - gu
  - ha
  - hi
  - ig
  - id
  - ja
  - rn
  - ko
  - ky
  - mr
  - ne
  - om
  - ps
  - fa
  - pcm
  - pt
  - pa
  - ru
  - gd
  - sr
  - si
  - so
  - es
  - sw
  - ta
  - te
  - th
  - ti
  - tr
  - uk
  - ur
  - uz
  - vi
  - cy
  - yo
size_categories:
  - 1M<n<10M
licenses:
  - cc-by-nc-sa-4.0
multilinguality:
  - multilingual
source_datasets:
  - original
paperswithcode_id: xl-sum
annotations_creators:
  - found
language_creators:
  - found
pretty_name: XL-Sum

Dataset Card for "XL-Sum"

Table of Contents

Dataset Description

Dataset Summary

We present XLSum, a comprehensive and diverse dataset comprising 1.35 million professionally annotated article-summary pairs from BBC, extracted using a set of carefully designed heuristics. The dataset covers 45 languages ranging from low to high-resource, for many of which no public dataset is currently available. XL-Sum is highly abstractive, concise, and of high quality, as indicated by human and intrinsic evaluation.

Supported Tasks and Leaderboards

Tasks: Summarization

Leaderboards: ExplainaBoard

Languages

  • amharic
  • arabic
  • azerbaijani
  • bengali
  • burmese
  • chinese_simplified
  • chinese_traditional
  • english
  • french
  • gujarati
  • hausa
  • hindi
  • igbo
  • indonesian
  • japanese
  • kirundi
  • korean
  • kyrgyz
  • marathi
  • nepali
  • oromo
  • pashto
  • persian
  • pidgin
  • portuguese
  • punjabi
  • russian
  • scottish_gaelic
  • serbian_cyrillic
  • serbian_latin
  • sinhala
  • somali
  • spanish
  • swahili
  • tamil
  • telugu
  • thai
  • tigrinya
  • turkish
  • ukrainian
  • urdu
  • uzbek
  • vietnamese
  • welsh
  • yoruba

Dataset Structure

Data Instances

One example from the English dataset is given below in JSON format. 

{
  "gem_id": "GEM-xlsum_english-train-1589",
  "url": "https://www.bbc.com/news/technology-17657859",
  "title": "Yahoo files e-book advert system patent applications",
  "summary": "Yahoo has signalled it is investigating e-book adverts as a way to stimulate its earnings.",
  "text": "Yahoo's patents suggest users could weigh the type of ads against the sizes of discount before purchase. It says in two US patent applications that ads for digital book readers have been \"less than optimal\" to date. The filings suggest that users could be offered titles at a variety of prices depending on the ads' prominence They add that the products shown could be determined by the type of book being read, or even the contents of a specific chapter, phrase or word. The paperwork was published by the US Patent and Trademark Office late last week and relates to work carried out at the firm's headquarters in Sunnyvale, California. \"Greater levels of advertising, which may be more valuable to an advertiser and potentially more distracting to an e-book reader, may warrant higher discounts,\" it states. Free books It suggests users could be offered ads as hyperlinks based within the book's text, in-laid text or even \"dynamic content\" such as video. Another idea suggests boxes at the bottom of a page could trail later chapters or quotes saying \"brought to you by Company A\". It adds that the more willing the customer is to see the ads, the greater the potential discount. \"Higher frequencies... may even be great enough to allow the e-book to be obtained for free,\" it states. The authors write that the type of ad could influence the value of the discount, with \"lower class advertising... such as teeth whitener advertisements\" offering a cheaper price than \"high\" or \"middle class\" adverts, for things like pizza. The inventors also suggest that ads could be linked to the mood or emotional state the reader is in as a they progress through a title. For example, they say if characters fall in love or show affection during a chapter, then ads for flowers or entertainment could be triggered. The patents also suggest this could applied to children's books - giving the Tom Hanks animated film Polar Express as an example. It says a scene showing a waiter giving the protagonists hot drinks \"may be an excellent opportunity to show an advertisement for hot cocoa, or a branded chocolate bar\". Another example states: \"If the setting includes young characters, a Coke advertisement could be provided, inviting the reader to enjoy a glass of Coke with his book, and providing a graphic of a cool glass.\" It adds that such targeting could be further enhanced by taking account of previous titles the owner has bought. 'Advertising-free zone' At present, several Amazon and Kobo e-book readers offer full-screen adverts when the device is switched off and show smaller ads on their menu screens, but the main text of the titles remains free of marketing. Yahoo does not currently provide ads to these devices, and a move into the area could boost its shrinking revenues. However, Philip Jones, deputy editor of the Bookseller magazine, said that the internet firm might struggle to get some of its ideas adopted. \"This has been mooted before and was fairly well decried,\" he said. \"Perhaps in a limited context it could work if the merchandise was strongly related to the title and was kept away from the text. \"But readers - particularly parents - like the fact that reading is an advertising-free zone. Authors would also want something to say about ads interrupting their narrative flow.\""
}

When downloading the dataset, the intended language name is required. For instance:

from datasets import load_dataset
ds = load_dataset("GEM/xlsum", "english")

Data Fields

  • gem_id: A string representing the article ID.
  • url: A string representing the article URL.
  • title: A string containing the article title.
  • summary: A string containing the article summary.
  • text : A string containing the article text.

Data Splits

We used a 80%-10%-10% split for all languages with a few exceptions. English was split 93%-3.5%-3.5% for the evaluation set size to resemble that of CNN/DM and XSum; Scottish Gaelic, Kyrgyz and Sinhala had relatively fewer samples, their evaluation sets were increased to 500 samples for more reliable evaluation. Same articles were used for evaluation in the two variants of Chinese and Serbian to prevent data leakage in multilingual training. Individual dataset download links with train-dev-test example counts are given below:

Language ISO 639-1 Code BBC subdomain(s) Train Dev Test Total
Amharic am https://www.bbc.com/amharic 5761 719 719 7199
Arabic ar https://www.bbc.com/arabic 37519 4689 4689 46897
Azerbaijani az https://www.bbc.com/azeri 6478 809 809 8096
Bengali bn https://www.bbc.com/bengali 8102 1012 1012 10126
Burmese my https://www.bbc.com/burmese 4569 570 570 5709
Chinese (Simplified) zh-CN https://www.bbc.com/ukchina/simp, https://www.bbc.com/zhongwen/simp 37362 4670 4670 46702
Chinese (Traditional) zh-TW https://www.bbc.com/ukchina/trad, https://www.bbc.com/zhongwen/trad 37373 4670 4670 46713
English en https://www.bbc.com/english, https://www.bbc.com/sinhala * 306522 11535 11535 329592
French fr https://www.bbc.com/afrique 8697 1086 1086 10869
Gujarati gu https://www.bbc.com/gujarati 9119 1139 1139 11397
Hausa ha https://www.bbc.com/hausa 6418 802 802 8022
Hindi hi https://www.bbc.com/hindi 70778 8847 8847 88472
Igbo ig https://www.bbc.com/igbo 4183 522 522 5227
Indonesian id https://www.bbc.com/indonesia 38242 4780 4780 47802
Japanese ja https://www.bbc.com/japanese 7113 889 889 8891
Kirundi rn https://www.bbc.com/gahuza 5746 718 718 7182
Korean ko https://www.bbc.com/korean 4407 550 550 5507
Kyrgyz ky https://www.bbc.com/kyrgyz 2266 500 500 3266
Marathi mr https://www.bbc.com/marathi 10903 1362 1362 13627
Nepali np https://www.bbc.com/nepali 5808 725 725 7258
Oromo om https://www.bbc.com/afaanoromoo 6063 757 757 7577
Pashto ps https://www.bbc.com/pashto 14353 1794 1794 17941
Persian fa https://www.bbc.com/persian 47251 5906 5906 59063
Pidgin** pcm https://www.bbc.com/pidgin 9208 1151 1151 11510
Portuguese pt https://www.bbc.com/portuguese 57402 7175 7175 71752
Punjabi pa https://www.bbc.com/punjabi 8215 1026 1026 10267
Russian ru https://www.bbc.com/russian, https://www.bbc.com/ukrainian * 62243 7780 7780 77803
Scottish Gaelic gd https://www.bbc.com/naidheachdan 1313 500 500 2313
Serbian (Cyrillic) sr https://www.bbc.com/serbian/cyr 7275 909 909 9093
Serbian (Latin) sr https://www.bbc.com/serbian/lat 7276 909 909 9094
Sinhala si https://www.bbc.com/sinhala 3249 500 500 4249
Somali so https://www.bbc.com/somali 5962 745 745 7452
Spanish es https://www.bbc.com/mundo 38110 4763 4763 47636
Swahili sw https://www.bbc.com/swahili 7898 987 987 9872
Tamil ta https://www.bbc.com/tamil 16222 2027 2027 20276
Telugu te https://www.bbc.com/telugu 10421 1302 1302 13025
Thai th https://www.bbc.com/thai 6616 826 826 8268
Tigrinya ti https://www.bbc.com/tigrinya 5451 681 681 6813
Turkish tr https://www.bbc.com/turkce 27176 3397 3397 33970
Ukrainian uk https://www.bbc.com/ukrainian 43201 5399 5399 53999
Urdu ur https://www.bbc.com/urdu 67665 8458 8458 84581
Uzbek uz https://www.bbc.com/uzbek 4728 590 590 5908
Vietnamese vi https://www.bbc.com/vietnamese 32111 4013 4013 40137
Welsh cy https://www.bbc.com/cymrufyw 9732 1216 1216 12164
Yoruba yo https://www.bbc.com/yoruba 6350 793 793 7936

* A lot of articles in BBC Sinhala and BBC Ukrainian were written in English and Russian respectively. They were identified using Fasttext and moved accordingly.

** West African Pidgin English

Dataset Creation

Curation Rationale

State-of-the-art text summarization models are heavily data-driven, i.e., a large number of article-summary pairs are required to train them effectively. As a result, abstractive summarization has centered around the English language, as most large abstractive summarization datasets are available in English only. Though there have been some recent efforts for curating multilingual abstractive summarization datasets, they are limited in terms of the number of languages covered, the number of training samples, or both. To this end, we curate XL-Sum, a large-scale abstractive summarization dataset of 1.35 million news articles from 45 languages crawled from the British Broadcasting Corporation website.

Source Data

BBC News

Initial Data Collection and Normalization

We designed a crawler to recursively crawl pages starting from the homepage by visiting different article links present in each page visited. We were able to take advantage of the fact that all BBC sites have somewhat similar structures, and were able to scrape articles from all sites. We discarded pages with no textual contents (mostly pages consisting of multimedia contents) before further processing. We designed a number of heuristics to make the extraction effective by carefully examining the HTML structures of the crawled pages:

  1. The desired summary must be present within the beginning two paragraphs of an article.
  2. The summary paragraph must have some portion of texts in bold format.
  3. The summary paragraph may contain some hyperlinks that may not be bold. The proportion of bold texts and hyperlinked texts to the total length of the paragraph in consideration must be at least 95%.
  4. All texts except the summary and the headline must be included in the input text (including image captions).
  5. The input text must be at least twice as large as the summary.

Who are the source language producers?

BBC News Editorial Team

Annotations

Annotation process

BBC typically provides a summary of a whole article in the form of a bold paragraph containing one or two sentences at the beginning of each article. These summaries are written professionally by the authors of the articles in order to convey its main story within one small paragraph. This is in contrast to the headline which serves to draw the attention of viewers into reading the article. We used the bold texts as summary and the rest of the article as input.

Who are the annotators?

BBC News Editorial Team

Personal and Sensitive Information

Meta-information like author names are discarded. However, we cannot guarantee removal of all personal information.

Considerations for Using the Data

Social Impact of Dataset

We believe that our efforts in this work will encourage the community to push the boundaries of abstractive text summarization beyond the English language, especially for low and mid-resource languages, bringing technological advances to communities of these languages that have been traditionally under-served.

Discussion of Biases

Human evaluation showed most languages had a high percentage of good summaries in the upper nineties, almost none of the summaries contained any conflicting information, while about one-third on average had information that was not directly inferrable from the source article.

Other Known Limitations

The dataset is limited to news domain only.

Additional Information

Dataset Curators

Authors of this paper

Licensing Information

Contents of this repository are restricted to only non-commercial research purposes under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). Copyright of the dataset contents belongs to the original copyright holders.

Citation Information

If you use any of the datasets, models or code modules, please cite the following paper:

@inproceedings{hasan-etal-2021-xl,
    title = "{XL}-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages",
    author = "Hasan, Tahmid  and
      Bhattacharjee, Abhik  and
      Islam, Md. Saiful  and
      Mubasshir, Kazi  and
      Li, Yuan-Fang  and
      Kang, Yong-Bin  and
      Rahman, M. Sohel  and
      Shahriyar, Rifat",
    booktitle = "Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.findings-acl.413",
    pages = "4693--4703",
}

Contributions

Thanks to @abhik1505040 and @Tahmid for adding this dataset.