This repository contains the code, data, and models of the paper titled "XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages" published in Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021.
- Check out CrossSum (ACL'23), the cross-lingual extension of XL-Sum!
- XL-Sum is now on ExplainaBoard! The leaderboard can be found here.
- XL-Sum can now be downloaded via HuggingFace Datasets.
- The multilingual checkpoint is also available at Huggingface Model Hub.
Disclaimer: You must agree to the license and terms of use before using the dataset.
We are releasing two versions of the dataset: an older version that has been reported in the paper; and a newer version with another added language (Traditional Chinese), more data, better formatting, better extraction, larger evaluation splits, and deduplication. We recommend using the latter and thus have organized the repository with data counts and benchmarks of the newer version. The new version contains a total of 1.35 million article-summary pairs, making XL-Sum the largest text summarization dataset publicly available.
All dataset files are in .jsonl format i.e. one JSON per line. One example from the English dataset is given below in JSON format. The fields are self-explanatory.
{
"id": "technology-17657859",
"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.\""
}
Download the complete dataset. See the legacy section for the older version(s).
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:
* 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
We are releasing a multilingual model checkpoint trained for 50k steps on the new data. To use this model for evaluation/inference refer to Training & Evaluation.
Multilingual model scores on test sets are given below. We are also releasing the model-generated outputs for future analysis.
| Language | ROUGE-1 / ROUGE-2 / ROUGE-L |
|---|---|
| Amharic | 20.0485 / 7.4111 / 18.0753 |
| Arabic | 34.9107 / 14.7937 / 29.1623 |
| Azerbaijani | 21.4227 / 9.5214 / 19.3331 |
| Bengali | 29.5653 / 12.1095 / 25.1315 |
| Burmese | 15.9626 / 5.1477 / 14.1819 |
| Chinese (Simplified) | 39.4071 / 17.7913 / 33.406 |
| Chinese (Traditional) | 37.1866 / 17.1432 / 31.6184 |
| English | 37.601 / 15.1536 / 29.8817 |
| French | 35.3398 / 16.1739 / 28.2041 |
| Gujarati | 21.9619 / 7.7417 / 19.86 |
| Hausa | 39.4375 / 17.6786 / 31.6667 |
| Hindi | 38.5882 / 16.8802 / 32.0132 |
| Igbo | 31.6148 / 10.1605 / 24.5309 |
| Indonesian | 37.0049 / 17.0181 / 30.7561 |
| Japanese | 48.1544 / 23.8482 / 37.3636 |
| Kirundi | 31.9907 / 14.3685 / 25.8305 |
| Korean | 23.6745 / 11.4478 / 22.3619 |
| Kyrgyz | 18.3751 / 7.9608 / 16.5033 |
| Marathi | 22.0141 / 9.5439 / 19.9208 |
| Nepali | 26.6547 / 10.2479 / 24.2847 |
| Oromo | 18.7025 / 6.1694 / 16.1862 |
| Pashto | 38.4743 / 15.5475 / 31.9065 |
| Persian | 36.9425 / 16.1934 / 30.0701 |
| Pidgin | 37.9574 / 15.1234 / 29.872 |
| Portuguese | 37.1676 / 15.9022 / 28.5586 |
| Punjabi | 30.6973 / 12.2058 / 25.515 |
| Russian | 32.2164 / 13.6386 / 26.1689 |
| Scottish Gaelic | 29.0231 / 10.9893 / 22.8814 |
| Serbian (Cyrillic) | 23.7841 / 7.9816 / 20.1379 |
| Serbian (Latin) | 21.6443 / 6.6573 / 18.2336 |
| Sinhala | 27.2901 / 13.3815 / 23.4699 |
| Somali | 31.5563 / 11.5818 / 24.2232 |
| Spanish | 31.5071 / 11.8767 / 24.0746 |
| Swahili | 37.6673 / 17.8534 / 30.9146 |
| Tamil | 24.3326 / 11.0553 / 22.0741 |
| Telugu | 19.8571 / 7.0337 / 17.6101 |
| Thai | 37.3951 / 17.275 / 28.8796 |
| Tigrinya | 25.321 / 8.0157 / 21.1729 |
| Turkish | 32.9304 / 15.5709 / 29.2622 |
| Ukrainian | 23.9908 / 10.1431 / 20.9199 |
| Urdu | 39.5579 / 18.3733 / 32.8442 |
| Uzbek | 16.8281 / 6.3406 / 15.4055 |
| Vietnamese | 32.8826 / 16.2247 / 26.0844 |
| Welsh | 32.6599 / 11.596 / 26.1164 |
| Yoruba | 31.6595 / 11.6599 / 25.0898 |
- See rouge module.
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.
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",
}