AraSeg: Arabic Sentence Segmentation

Looking for AraSeg? This page is a pointer. The dataset and models live in the repositories below.

👉 Collection: Arabic Sentence Segmentation

AraSeg is a manually annotated, genre-diverse benchmark for Arabic sentence segmentation, introduced in Arabic Sentence Segmentation Across Genres and Punctuation Conditions (Findings of EMNLP 2026).

Dataset

Dataset	Description
MBZUAI/AraSEG	2,187 documents, 121,623 sentences, 8 genres; four subsets: PA, NP, NoPnx-PA, NoPnx-NP

Models

CAMeLBERT-MSA fine-tuned for sentence segmentation, one model per task variant:

Model	Paragraph boundaries	Punctuation
MBZUAI/sentseg-camelbert-PA	✓	✓
MBZUAI/sentseg-camelbert-NP	✗	✓
MBZUAI/sentseg-camelbert-NoPnx-PA	✓	✗
MBZUAI/sentseg-camelbert-NoPnx-NP	✗	✗

Citation

@inproceedings{elkholy-etal-2026-arabic-sentence-segmentation,
    title = "{A}rabic Sentence Segmentation Across Genres and Punctuation Conditions",
    author = "Elkholy, Mohammed and
      Elmadani, Khalid N. and
      Habash, Nizar and
      Alhafni, Bashar",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2026",
    month = oct,
    year = "2026",
    address = "Budapest, Hungary",
    publisher = "Association for Computational Linguistics"
}