Elamur Research⌕
☰
home / papers / 2023 / LANS: Large-scale Arabic News Summarization Corpus

LANS: Large-scale Arabic News Summarization Corpus

Paper · 2023 · Abdulaziz Alhamadani, Xuchao Zhang, Jianfeng He et al.

Summary

Arabic Text Summarization (ATS) lags behind English largely because existing corpora are tiny or homogeneous. They hold at most about 41K pairs and usually come from one or a few outlets. LANS addresses this by scraping 22 established newspapers from 19 Arab countries over 1999–2019. Each article is paired with the summary its journalist wrote, which the authors harvest from the page's HTML og:description metadata field. After cleaning and filtering roughly 11M scraped items, 8.4M article–summary pairs remain. This makes LANS by far the largest ATS resource. It is also diverse in two ways: it reflects regional variation in Modern Standard Arabic shaped by local dialects, and each source covers at least 7 news categories. Quality was checked with automatic measures of abstractiveness and conciseness and with a blind expert study on 1,000 samples. In that study, the metadata summary was preferred in 95.4% of cases. The work matters because it gives Arabic summarization enough data to train modern models.

Key points

  • LANS has 8,443,484 article–summary pairs from 22 newspapers in 19 countries, covering 1999–2019. The largest contributors are Ahram Gate (about 2.04M) and Alryiadh (about 1.0M). The dataset is available on request from the first author.
  • Sources were chosen by three explicit criteria: a deep electronic archive, at least 7 topic categories, and an author summary present in the page metadata. Websites of TV-channel outlets such as Aljazeera and Arabic BBC were excluded.
  • Each source was profiled with coverage, density and compression scores. These vary widely between outlets, for example density ranges from 0.30 for WAL to 1.01 for Hespress, which signals both diversity and a fair degree of abstractiveness.
  • In the human study, native-speaker experts with Arabic-language degrees compared the LANS summary with an mT5-generated summary, plus a "none" option. They picked LANS 954 times out of 1,000 and chose "none" only twice.
  • The authors acknowledge one limitation: the data is unevenly spread across countries, because media landscapes differ in size.

Interesting ideas

  • I1. In the human check, the candidate outputs were shuffled and anonymized, and judges also had a "neither is acceptable" option, which reduces position bias and exposes cases where every candidate fails.
  • I2. The authors profile each source separately with copy-oriented statistics (coverage, density, compression) instead of one corpus-wide average, which shows how sources differ in style.
  • I3. Generated summaries scored low on ROUGE against LANS references, yet experts strongly preferred the references, a reminder that overlap metrics can misrepresent quality.
  • I4. Sources were selected on purpose to cover regional variation within one standard language, since local dialects color its vocabulary and style.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • Author-written metadata such as og:description acts as free, large-scale supervision that comes from the publisher rather than from annotators, and the same trick works for any web genre with descriptive meta tags. — (back): we could harvest description fields of English pages as reference summaries to evaluate the text → graph → text compressor on what a short rendering should keep.
  • I1 (learn): when an LLM teacher or human adjudicates between witness parsers' outputs, hide the origin, randomize order, and include a "neither" option as a built-in negative control.
  • I2 (back): computing these statistics on generated text versus its input would show how much generation truly rebuilds from the meaning graph rather than copying surface fragments.
  • I3 (answer): grade free-text answers by checking meaning against the world of the text rather than by word overlap, so that paraphrases are not penalized.
  • I4 (tree): build evaluation sets from regionally diverse English outlets (British, Indian, Nigerian, etc.) to test parser robustness on lexical and stylistic variation of standard English.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢MicrosoftRedmond · United States378 works · 161,630 citations🎓Virginia TechBlacksburg · United States39 works · 43,418 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 3,066 worksSummarization · 243 works

Cite

Cite as Abdulaziz Alhamadani, Xuchao Zhang, Jianfeng He, Aadyant Khatri and Chang-Tien Lu. 2023. LANS: Large-scale Arabic News Summarization Corpus. Proceedings of ArabicNLP 2023.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2023.arabicnlp-1.8

@inproceedings{alhamadani2023lans,
  title = {{LANS: Large-scale Arabic News Summarization Corpus}},
  author = {Abdulaziz Alhamadani, Xuchao Zhang, Jianfeng He, Aadyant Khatri and Chang-Tien Lu},
  booktitle = {Proceedings of ArabicNLP 2023},
  year = {2023},
  doi = {10.18653/v1/2023.arabicnlp-1.8},
  url = {https://doi.org/10.18653/v1/2023.arabicnlp-1.8}
}

References

40 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Fouad Soufiane Douzidia and Guy Lapalme. 2004. Lakhas, an arabic summarization system. In Proceedings of DUC, volume 4, pages 128–135. find
  2. Citeseer. 2022. Gate ahram newspaper (egypt). http://gate.ahram.org.eg/. Accessed: 2020-02-02. ↗
  3. Mahmoud El-Defrawy, Yasser El-Sonbaty, and Nahla Belal. 2015. Enhancing root extractors using light stemmers. In Proceedings of the 29th Pacific Asia Conference on Language, Information and Computation: Posters, pages 157–166. find
  4. Ahmed Abdelali, Kareem Darwish, Nadir Durrani, and Hamdy Mubarak. 2016. Farasa: A fast and furious segmenter for arabic. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, pages 11–16.
  5. Fatma El-Ghannam and Tarek El-Shishtawy. 2014. Multi-topic multi-document summarizer. arXiv preprint arXiv:1401.0640. ↗
  6. Abdullah M Abu Nada, Eman Alajrami, Ahmed A Al-Saqqa, and Samy S Abu-Naser. 2020. Arabic text summarization using arabert model using extractive text summarization approach. International Journal of Academic Information Systems Research (IJAISR). find
  7. Mahmoud El-Haj and Rim Koulali. 2013. Kalimat a multipurpose arabic corpus. In Second Workshop on Arabic Corpus Linguistics (WACL-2), pages 22–25. find
  8. Fatima T AL-Khawaldeh and Venus W Samawi. 2015. Lexical cohesion and entailment based segmentation for arabic text summarization (lceas). World of Computer Science & Information Technology Journal, 5(3). find
  9. Mahmoud El-Haj, Udo Kruschwitz, and Chris Fox. 2010. Using mechanical turk to create a corpus of arabic summaries. find
  10. Molham Al-Maleh and Said Desouki. 2020. Arabic text summarization using deep learning approach. Journal of Big Data, 7(1):1–17. ↗
  11. Wafaa S El-Kassas, Cherif R Salama, Ahmed A Rafea, and Hoda K Mohamed. 2021. Automatic text summarization: A comprehensive survey. Expert Systems with Applications, 165:113679. ↗
  12. Asma Bader Al-Saleh and Mohamed El Bachir Menai. 2016. Automatic arabic text summarization: a survey. Artificial Intelligence Review, 45(2):203–234. ↗
  13. Ibrahim Abu El-Khair. 2017. Effects of stop words elimination for arabic information retrieval: a comparative study. arXiv preprint arXiv:1702.01925. ↗
  14. Ahmad T Al-Taani and Maha M Al-Omour. 2014. An extractive graph-based arabic text summarization approach. In The International Arab Conference on Information Technology. find
  15. Asmaa Elsaid, Ammar Mohammed, Lamiaa Fattouh, and Mohamed Sakre. 2022. A comprehensive review of arabic text summarization. IEEE Access. ↗
  16. Abdullah Ayedh, Guanzheng Tan, Khaled Alwesabi, and Hamdi Rajeh. 2016. The effect of preprocessing on arabic document categorization. Algorithms, 9(2):27. ↗
  17. Hamzah Noori Fejer and Nazlia Omar. 2014. Automatic arabic text summarization using clustering and keyphrase extraction. In Proceedings of the 6th International Conference on Information Technology and Multimedia, pages 293–298. IEEE. find
  18. Riadh Belkebir and Ahmed Guessoum. 2015. A supervised approach to arabic text summarization using adaboost. In New contributions in information systems and technologies, pages 227–236. Springer. ↗
  19. Kavita Ganesan. 2018. Rouge 2.0: Updated and improved measures for evaluation of summarization tasks. arXiv preprint arXiv:1803.01937. ↗
  20. Max Grusky, Mor Naaman, and Yoav Artzi. 2018. Newsroom: A dataset of 1.3 million summaries with diverse extractive strategies. arXiv preprint arXiv:1804.11283. ↗
  21. Rishi Bommasani and Claire Cardie. 2020. Intrinsic evaluation of summarization datasets. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 8075–8096. find
  22. Tahmid Hasan, Abhik Bhattacharjee, Md Saiful Islam, Kazi Samin, Yuan-Fang Li, Yong-Bin Kang, M Sohel Rahman, and Rifat Shahriyar. 2021. Xl-sum: Large-scale multilingual abstractive summarization for 44 languages. arXiv preprint arXiv:2106.13822. ↗
  23. Mohamed Boudchiche and Azzeddine Mazroui. 2019. A hybrid approach for arabic lemmatization. International Journal of Speech Technology, 22(3):563–573. ↗
  24. Amina Chouigui, Oussama Ben Khiroun, and Bilel Elayeb. 2021. An arabic multi-source news corpus: Experimenting on single-document extractive summarization. Arabian Journal for Science and Engineering, 46(4):3925–3938. ↗
  25. Marcin Junczys-Dowmunt, Roman Grundkiewicz, Tomasz Dwojak, Hieu Hoang, Kenneth Heafield, Tom Neckermann, Frank Seide, Ulrich Germann, Alham Fikri Aji, Nikolay Bogoychev, André F. T. Martins, and Alexandra Birch. 2018. Marian: Fast neural machine translation in C++. In Proceedings of ACL 2018, System Demonstrations, pages 116–121, Melbourne, Australia. Association for Computational Linguistics.
  26. Zi-Yi Dou, Pengfei Liu, Hiroaki Hayashi, Zhengbao Jiang, and Graham Neubig. 2020. Gsum: A general framework for guided neural abstractive summarization. arXiv preprint arXiv:2010.08014. ↗
  27. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov, and Luke Zettlemoyer. 2019. Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. arXiv preprint arXiv:1910.13461. ↗
  28. Michael Völske, Martin Potthast, Shahbaz Syed, and Benno Stein. 2017. TL;DR: Mining Reddit to learn automatic summarization. In Proceedings of the Workshop on New Frontiers in Summarization, pages 59–63, Copenhagen, Denmark. Association for Computational Linguistics.
  29. Edward Loper and Steven Bird. 2002. Nltk: The natural language toolkit. In In Proceedings of the ACL Workshop on Effective Tools and Methodologies for Teaching Natural Language Processing and Computational Linguistics. Philadelphia: Association for Computational Linguistics. ↗
  30. Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, and Colin Raffel. 2020. mt5: A massively multilingual pre-trained text-to-text transformer. arXiv preprint arXiv:2010.11934. ↗
  31. Jingqing Zhang, Yao Zhao, Mohammad Saleh, and Peter Liu. 2020. Pegasus: Pre-training with extracted gap-sentences for abstractive summarization. In International Conference on Machine Learning, pages 11328–11339. PMLR. ↗
  32. Hamdy Mubarak. 2017. Build fast and accurate lemmatization for arabic. arXiv preprint arXiv:1710.06700. ↗
  33. Ossama Obeid, Nasser Zalmout, Salam Khalifa, Dima Taji, Mai Oudah, Bashar Alhafni, Go Inoue, Fadhl Eryani, Alexander Erdmann, and Nizar Habash. 2020. Camel tools: An open source python toolkit for arabic natural language processing. In Proceedings of the 12th language resources and evaluation conference, pages 7022–7032.
  34. Arfath Pasha, Mohamed Al-Badrashiny, Mona T Diab, Ahmed El Kholy, Ramy Eskander, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. Madamira: A fast, comprehensive tool for morphological analysis and disambiguation of arabic. In LREC, volume 14, pages 1094–1101.
  35. Dragomir Radev, Eduard Hovy, and Kathleen McKeown. 2002. Introduction to the special issue on summarization. Computational linguistics, 28(4):399– 408. ↗
  36. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2019. Exploring the limits of transfer learning with a unified text-to-text transformer. arXiv preprint arXiv:1910.10683. ↗
  37. Abigail See, Peter J Liu, and Christopher D Manning. 2017. Get to the point: Summarization with pointer-generator networks. arXiv preprint arXiv:1704.04368. ↗
  38. Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, and Angela Fan. 2020. Multilingual translation with extensible multilingual pretraining and finetuning. arXiv preprint arXiv:2008.00401. ↗
  39. Jörg Tiedemann and Santhosh Thottingal. 2020. OPUS-MT – building open translation services for the world. In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation, pages 479–480, Lisboa, Portugal. European Association for Machine Translation. find
  40. Chau Tran, Yuqing Tang, Xian Li, and Jiatao Gu. 2020. Cross-lingual retrieval for iterative self-supervised training. Advances in Neural Information Processing Systems, 33:2207–2219. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/alhamadani-2023-lans-large-scale-arabic-news/