Elamur Research⌕
☰
home / papers / 2025 / AraFinNews: Arabic Financial Summarisation with Domain-Adapted LLMs

AraFinNews: Arabic Financial Summarisation with Domain-Adapted LLMs

Summary

Arabic has far fewer domain-specific NLP resources than English, and financial text is especially underserved. Existing Arabic summarisation corpora mostly cover general news. The one earlier financial dataset, FinAraSum, is relatively small and was never released. To fill this gap, the authors build AraFinNews, an openly available collection of 212,500 Arabic financial articles. Each article is paired with its headline, and the headline serves as a short abstractive summary. The material comes from Argaam.com, a Gulf financial news platform, and covers about ten years of reporting. The authors present it as an Arabic analogue of CNN/DailyMail.

They then fine-tune and compare three text-to-text transformers: multilingual mT5, general Arabic AraT5, and FinAraT5, which received extra pretraining on financial Arabic. The domain-adapted model produces more coherent summaries and is better at preserving numbers and named entities. The work supports the argument that domain-specific pretraining matters for professional Arabic text generation. It also provides a reproducible benchmark for this task and for related tasks such as NER, event extraction and sentiment analysis.

Key points

  • AraFinNews contains 212,500 article–headline pairs dated from 8 October 2015 to 29 July 2025. It is released for non-commercial research on GitHub (ArabicNLP-UK/AraFinNews).
  • The source is Argaam.com, which also aggregates world financial coverage from outlets such as the Financial Times, Forbes, Bloomberg and the World Gold Council.
  • The content spans corporate disclosures, market analyses and macroeconomic reporting.
  • The compared models are mT5, AraT5 and the domain-adapted FinAraT5. Evaluation looks at accuracy, numerical reliability and closeness to professional reporting style.
  • Prior Arabic summarisation resources (Kalimat, EASC, XLSum, WikiLingua, SumArabic, HASD) are general-domain. The earlier financial set FinAraSum has roughly 44,900 articles and is not public.

Interesting ideas

  • I1. Using professionally written headlines as naturally occurring abstractive summaries yields a large supervised dataset without any annotation cost.
  • I2. Scoring summaries on whether numbers and named entities survive, not only on overall overlap, separates fluent output from faithful output.
  • I3. Continued pretraining on in-domain text (FinAraT5 over AraT5) gives measurable gains on domain tasks, so adapting the domain can matter as much as model choice.
  • I4. A corpus that spans a decade captures how reporting style and vocabulary drift over time, which makes temporal splits possible.
  • I5. Financial news is entity-centric and full of quantities, which makes it a natural testbed for tracking who holds what and how values change.

For Mova

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

  • I1 (back): News headline–body pairs could serve as free gold data for testing whether the text → graph → text compressor keeps the core event of an article.
  • I2 (learn): Evaluations could include a dedicated check that quantities and entities in generated or answered text match the source, which works as a targeted negative control for hallucination.
  • I3 (tree): gparse could be further pretrained or fine-tuned on financial or other specialist English before parsing such text, to cope with dense numeric and terminological constructions.
  • I4 (learn): Silver data and test sets could be split by date to check that parsers and readers do not just memorise the style of a single period.
  • I5 (world): Financial articles could stress-test the world model's handling of holdings, measures, values and event order, with headlines as a compact check of what the world model should contain.

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

Fields: Linguistics & languages

Topics Language models & pretraining Summarization Semitic languages

 Institutions

🎓Lancaster UniversityLancaster · United Kingdom13 works · 1,069 citations🎓VinUniversityHanoi · Vietnam3 works · 43 citations

Cite

Cite as Mo El-Haj and Paul Rayson. 2025. AraFinNews: Arabic Financial Summarisation with Domain-Adapted LLMs. arXiv:2511.01265 · IEEE BigData 2025.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2511.01265 · arXiv 2511.01265

@misc{elhaj2025arafinnews,
  title = {{AraFinNews: Arabic Financial Summarisation with Domain-Adapted LLMs}},
  author = {Mo El-Haj and Paul Rayson},
  year = {2025},
  doi = {10.48550/arxiv.2511.01265},
  eprint = {2511.01265},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2511.01265}
}

References

49 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Statista Research Department, “The most spoken languages worldwide,” 2024, accessed July 29, 2025. [Online]. Available: https://www.statista.com/statistics/266808/the-most-spoken-languages-worldwide/ ↗
  2. Using mechanical turk to create a corpus of arabic summaries. European Language Resources Association, 2010. find
  3. M. El-Haj and P. Rayson, “OSMAN — a novel Arabic readability metric,” in Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16). Portorož, Slovenia: European Language Resources Association (ELRA), May 2016, pp. 250–255. [Online]. Available: https://aclanthology.org/L16-1038 ↗ our card
  4. W. Antoun, F. Baly, and H. Hajj, “Arabert: Transformer-based model for arabic language understanding,” in LREC 2020 Workshop Language Resources and Evaluation Conference 11–16 May 2020. Marseille, France: European Language Resource Association, 2020, p. 9. [Online]. Available: https://aclanthology.org/2020.osact-1.2 ↗ our card
  5. M. Abdul-Mageed, A. Elmadany, and E. M. B. Nagoudi, “ARBERT & MARBERT: Deep Bidirectional Transformers for Arabic,” in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Online: Association for Computational Linguistics, Aug. 2021, pp. 7088–7105. [Online]. Available: https://aclanthology.org/2021.acl-long.551 ↗ our card
  6. M. K. Eddine, N. Tomeh, N. Habash, J. L. Roux, and M. Vazirgiannis, “AraBART: a Pretrained Arabic Sequence-to-Sequence Model for Abstractive Summarization,” arXiv:2203.10945 [cs], Mar. 2022, arXiv: 2203.10945. [Online]. Available: http://arxiv.org/abs/2203.10945 ↗
  7. E. M. B. Nagoudi, A. Elmadany, and M. Abdul-Mageed, “AraT5: Text-to-text transformers for Arabic language generation,” in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Dublin, Ireland: Association for Computational Linguistics, May 2022, pp. 628–647. [Online]. Available: https://aclanthology.org/2022.acl-long.47 ↗ our card
  8. Y. Yang, M. C. S. UY, and A. Huang, “Finbert: A pretrained language model for financial communications,” 2020. find
  9. J. Lee, W. Yoon, S. Kim, D. Kim, S. Kim, C. H. So, and J. Kang, “Biobert: a pre-trained biomedical language representation model for biomedical text mining,” Bioinformatics, vol. 36, no. 4, pp. 1234–1240, 2020. our card
  10. N. Zmandar, A. Singh, M. El-Haj, and P. Rayson, “Joint abstractive and extractive method for long financial document summarization,” in Proceedings of the 3rd Financial Narrative Processing Workshop, 2021, pp. 99–105. find
  11. N. Zmandar, M. El-Haj, and P. Rayson, “Multilingual financial word embeddings for arabic, english and french,” in 2021 IEEE International Conference on Big Data (Big Data). IEEE, 2021, pp. 4584–4589. find
  12. M. El-Haj, M. Litvak, N. Pittaras, G. Giannakopoulos et al., “The financial narrative summarisation shared task (fns 2020),” in Proceedings of the 1st Joint Workshop on Financial Narrative Processing and MultiLing Financial Summarisation, 2020, pp. 1–12. find
  13. N. Zmandar, M. El-Haj, P. Rayson, M. Litvak, G. Giannakopoulos, N. Pittaras et al., “The financial narrative summarisation shared task fns 2021,” in Proceedings of the 3rd Financial Narrative Processing Workshop, 2021, pp. 120–125. find
  14. E. Zavitsanos, A. Kosmopoulos, G. Giannakopoulos, M. Litvak, B. Carbajo-Coronado, A. Moreno-Sandoval, and M. El-Haj, “The financial narrative summarisation shared task (fns 2023),” in 2023 IEEE International Conference on Big Data (BigData). IEEE, 2023, pp. 2890–2896. find
  15. K. Filippova, M. Surdeanu, M. Ciaramita, and H. Zaragoza, “Company-oriented extractive summarization of financial news,” in Proceedings of the 12th Conference of the European Chapter of the ACL (EACL 2009), 2009, pp. 246–254. find
  16. A. L. Berger and V. O. Mittal, “Ocelot: a system for summarizing web pages,” in Proceedings of the 23rd annual international ACM SIGIR conference on Research and development in information retrieval, 2000, pp. 144–151. find
  17. K. Lopyrev, “Generating news headlines with recurrent neural networks,” arXiv preprint arXiv:1512.01712, 2015. ↗
  18. A. See, P. J. Liu, and C. D. Manning, “Get to the point: Summarization with pointer-generator networks,” arXiv preprint arXiv:1704.04368, 2017. ↗
  19. J. Cheng and M. Lapata, “Neural summarization by extracting sentences and words,” arXiv preprint arXiv:1603.07252, 2016. ↗ our card
  20. C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y. Zhou, W. Li, and P. J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,” Journal of machine learning research, vol. 21, no. 140, pp. 1–67, 2020. our card
  21. L. Xue, N. Constant, A. Roberts, M. Kale, R. Al-Rfou, A. Siddhant, A. Barua, and C. Raffel, “mT5: A massively multilingual pre-trained text-to-text transformer,” in Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Online: Association for Computational Linguistics, Jun. 2021, pp. 483–498. [Online]. Available: https://aclanthology.org/2021.naacl-main.41 ↗ our card
  22. M. Abdul-Mageed, A. Elmadany, A. Inciarte, M. T. I. Khondaker et al., “Jasmine: Arabic gpt models for few-shot learning,” in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 16 721–16 744. find
  23. M. Kahla, A. Novák, and Z. G. Yang, “Fine-tuning and multilingual pre-training for abstractive summarization task for the arabic language,” in Annales Mathematicae et Informaticae, vol. 57, 2023, pp. 24–16. find
  24. K. Morsi, W. Hussein, R. Ismail et al., “Arabic text summarization using transformer-based architectures.” International Journal for Computers & Their Applications, vol. 32, no. 1, 2025. find
  25. N. Essa, M. El-Gayar, and E. M. El-Daydamony, “Enhanced model for abstractive arabic text summarization using natural language generation and named entity recognition,” Neural Computing and Applications, vol. 37, no. 10, pp. 7279–7301, 2025. find
  26. A. Elsaid, “Hybrid arabic text summarization approach based on seq-to-seq and transformer,” Journal of the ACS Advances in Computer Science, vol. 16, no. 1, 2025. find
  27. N. Zmandar, M. El-Haj, and P. Rayson, “Finarat5: A text to text model for financial arabic text understanding and generation,” in Proceedings of the 4th Conference on Language, Data and Knowledge, 2023, pp. 262–273. find
  28. M. El-Haj and R. Koulali, “Kalimat a multipurpose arabic corpus,” in The Second Workshop on Arabic Corpus Linguistics (WACL-2), 2013, pp. 1–4. [Online]. Available: https://eprints.lancs.ac.uk/id/eprint/71282/ 1/KALIMAT_ELHAJ_KOULALI.pdf ↗
  29. R. Parker, D. Graff, K. Chen, J. Kong, and K. Maeda, “Arabic gigaword dataset,” 2009. [Online]. Available: https://catalog.ldc.upenn.edu/LDC2009T30 ↗
  30. T. Hasan, A. Bhattacharjee, M. S. Islam, K. Mubasshir, Y.-F. Li, Y.-B. Kang, M. S. Rahman, and R. Shahriyar, “XL-sum: Large-scale multilingual abstractive summarization for 44 languages,” in Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021. Online: Association for Computational Linguistics, Aug. 2021, pp. 4693–4703. [Online]. Available: https://aclanthology.org/2021.findings-acl.413 ↗
  31. F. Ladhak, E. Durmus, C. Cardie, and K. McKeown, “WikiLingua: A new benchmark dataset for cross-lingual abstractive summarization,” in Findings of the Association for Computational Linguistics: EMNLP 2020. Online: Association for Computational Linguistics, Nov. 2020, pp. 4034–4048. [Online]. Available: https://aclanthology.org/ 2020.findings-emnlp.360 ↗ our card
  32. M. Bani-Almarjeh and M.-B. Kurdy, “Arabic abstractive text summarization using rnn-based and transformer-based architectures,” Information Processing and Management, vol. 60, no. 2, p. 103227, 2023. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/S0306457322003284 ↗
  33. A. Elsaid, A. Mohammed, L. Fattouh, and M. Sakre, “Abstractive arabic text summarization based on mt5 and arabart transformers,” in 2023 Intelligent Methods, Systems, and Applications (IMSA), 2023, pp. 7–12. find
  34. M. EL-Haj, U. Kruschwitz, and C. Fox, “Multi-document arabic text summarisation,” in 2011 3rd Computer Science and Electronic Engineering Conference, CEEC’11, 07 2011, pp. 40–44. find
  35. F. S. Douzidia and G. Lapalme, “Lakhas, an arabic summarization system,” in 4th Document Understanding Conferences, 2004, pp. 128–135. [Online]. Available: https://api.semanticscholar.org/CorpusID: 18307049 ↗
  36. M. El-Haj and P. Rayson, “Using a keyness metric for single and multi document summarisation,” in Proceedings of the MultiLing 2013 Workshop on Multilingual Multi-document Summarization, 2013, pp. 64–71. find
  37. R. Koulali, M. El-Haj, and A. Meziane, “Arabic topic detection using automatic text summarisation,” in 2013 ACS International Conference on Computer Systems and Applications (AICCSA). IEEE, 2013, pp. 1–4. find
  38. M. Kahla, Z. G. Yang, and A. Novák, “Cross-lingual fine-tuning for abstractive Arabic text summarization,” in Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021). Held Online: INCOMA Ltd., Sep. 2021, pp. 655–663. [Online]. Available: https://aclanthology.org/2021.ranlp-1.74 ↗
  39. A. Mohammed, Z. Fuxi, M. Alghaili, E. Refaee, and M. Bamiah, “Borsah: An arabic sentiment financial tweets corpus,” in Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), M. El-Haj, P. Rayson, and A. Moore, Eds. Paris, France: European Language Resources Association (ELRA), 2018, pp. 17–22. find
  40. M. Kamal Eddine, A. Tixier, and M. Vazirgiannis, “BARThez: a skilled pretrained French sequence-to-sequence model,” in Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. Online and Punta Cana, Dominican Republic: Association for Computational Linguistics, Nov. 2021, pp. 9369–9390. [Online]. Available: https://aclanthology.org/2021.emnlp-main.740 ↗
  41. T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell et al., “Language models are few-shot learners,” Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020. our card
  42. A. Chowdhery, S. Narang, J. Devlin, M. Bosma, G. Mishra, A. Roberts, P. Barham, H. W. Chung, C. Sutton, S. Gehrmann et al., “Palm: Scaling language modeling with pathways,” Journal of Machine Learning Research, vol. 24, no. 240, pp. 1–113, 2023. our card
  43. H. W. Chung, L. Hou, S. Longpre, B. Zoph, Y. Tay, W. Fedus, Y. Li, X. Wang, M. Dehghani, S. Brahma et al., “Scaling instruction-finetuned language models,” Journal of Machine Learning Research, vol. 25, no. 70, pp. 1–53, 2024. our card
  44. W. Antoun, F. Baly, and H. Hajj, “Arabert: Transformer-based model for arabic language understanding,” arXiv preprint arXiv:2003.00104, 2020. ↗ our card
  45. S. Rothe, S. Narayan, and A. Severyn, “Leveraging pre-trained checkpoints for sequence generation tasks,” Transactions of the Association for Computational Linguistics, vol. 8, pp. 264–280, 2020. [Online]. Available: https://aclanthology.org/2020.tacl-1.18 ↗ our card
  46. C.-Y. Lin, “ROUGE: A Package for Automatic Evaluation of Summaries,” in Text Summarization Branches Out. Barcelona, Spain: Association for Computational Linguistics, Jul. 2004, pp. 74–81. [Online]. Available: https://aclanthology.org/W04-1013 ↗ our card
  47. S. Banerjee and A. Lavie, “Meteor: An automatic metric for mt evaluation with improved correlation with human judgments,” in Proceedings of the acl workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, 2005, pp. 65–72. our card
  48. M. Kamal Eddine, G. Shang, A. Tixier, and M. Vazirgiannis, “FrugalScore: Learning cheaper, lighter and faster evaluation metrics for automatic text generation,” in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Dublin, Ireland: Association for Computational Linguistics, May 2022, pp. 1305–1318. [Online]. Available: https://aclanthology.org/2022.acl-long.93 ↗
  49. T. Sellam, D. Das, and A. P. Parikh, “Bleurt: Learning robust metrics for text generation,” arXiv preprint arXiv:2004.04696, 2020. ↗

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/el-haj-2025-arafinnews-arabic-financial-summarisation-domain/