ASAS (أساس, “foundation”) is a high-quality Arabic summarization dataset with sentence-level human validation and explicit supporting evidence from the source text.It is designed for long-form, factual, and faithful summarization research.
✅ Long source articles (news, essays, legal, science, general writing)
✅ Multi-sentence summaries… See the full description on the dataset page:
https://huggingface.co/datasets/HebArabNlpProject/ASAS.
