@inproceedings{rabih-etal-2026-benchmarking,
title = "Benchmarking Fine-Grained Readability Explanations for {A}rabic",
author = "Rabih, Nour and
Elmadani, Khalid N. and
Taha, Hanada and
Habash, Nizar",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-long.156/",
pages = "3445--3472",
ISBN = "979-8-89176-439-2",
abstract = "In this work, we introduce BAREC-X, the first explanation-annotated Arabic readability dataset, comprising 1,793 sentences drawn from the Balanced Arabic Readability Evaluation Corpus (BAREC) and labeled by expert annotators with linguistic justifications aligned with the BAREC guidelines. Annotators reached an average inter-annotator agreement of 93.3{\%} under the At-Least-1 metric (73.9{\%} Exact Match at the Category+Reason level, and 81.9{\%} at the Category level). We frame readability explanation as a gold-level-conditioned classification task: given a sentence and its gold readability level, a system must return the linguistic reasons that justify the label. We present the first systematic benchmark for this task on Arabic. We evaluate two contrasting paradigms: (i) eight LLMs under zero- and few-shot prompting, and (ii) a fully interpretable, training-free rule-based system grounded in 67 linguistic features spanning morphology, syntax, vocabulary, and syllabic structure. Gemini{~}3 and GPT-5.5 achieve the highest exact-match agreement on fine-grained reason prediction (55.8{\%} and 56.4{\%} Category+Reason), while the rule-based system achieves the strongest exact-match performance at the Category level (65.9{\%}). Together, BAREC-X and our benchmark establish a transparent foundation for interpretable Arabic readability explanation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rabih-etal-2026-benchmarking">
<titleInfo>
<title>Benchmarking Fine-Grained Readability Explanations for Arabic</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nour</namePart>
<namePart type="family">Rabih</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Khalid</namePart>
<namePart type="given">N</namePart>
<namePart type="family">Elmadani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hanada</namePart>
<namePart type="family">Taha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nizar</namePart>
<namePart type="family">Habash</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-439-2</identifier>
</relatedItem>
<abstract>In this work, we introduce BAREC-X, the first explanation-annotated Arabic readability dataset, comprising 1,793 sentences drawn from the Balanced Arabic Readability Evaluation Corpus (BAREC) and labeled by expert annotators with linguistic justifications aligned with the BAREC guidelines. Annotators reached an average inter-annotator agreement of 93.3% under the At-Least-1 metric (73.9% Exact Match at the Category+Reason level, and 81.9% at the Category level). We frame readability explanation as a gold-level-conditioned classification task: given a sentence and its gold readability level, a system must return the linguistic reasons that justify the label. We present the first systematic benchmark for this task on Arabic. We evaluate two contrasting paradigms: (i) eight LLMs under zero- and few-shot prompting, and (ii) a fully interpretable, training-free rule-based system grounded in 67 linguistic features spanning morphology, syntax, vocabulary, and syllabic structure. Gemini 3 and GPT-5.5 achieve the highest exact-match agreement on fine-grained reason prediction (55.8% and 56.4% Category+Reason), while the rule-based system achieves the strongest exact-match performance at the Category level (65.9%). Together, BAREC-X and our benchmark establish a transparent foundation for interpretable Arabic readability explanation.</abstract>
<identifier type="citekey">rabih-etal-2026-benchmarking</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.156/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>3445</start>
<end>3472</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Benchmarking Fine-Grained Readability Explanations for Arabic
%A Rabih, Nour
%A Elmadani, Khalid N.
%A Taha, Hanada
%A Habash, Nizar
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F rabih-etal-2026-benchmarking
%X In this work, we introduce BAREC-X, the first explanation-annotated Arabic readability dataset, comprising 1,793 sentences drawn from the Balanced Arabic Readability Evaluation Corpus (BAREC) and labeled by expert annotators with linguistic justifications aligned with the BAREC guidelines. Annotators reached an average inter-annotator agreement of 93.3% under the At-Least-1 metric (73.9% Exact Match at the Category+Reason level, and 81.9% at the Category level). We frame readability explanation as a gold-level-conditioned classification task: given a sentence and its gold readability level, a system must return the linguistic reasons that justify the label. We present the first systematic benchmark for this task on Arabic. We evaluate two contrasting paradigms: (i) eight LLMs under zero- and few-shot prompting, and (ii) a fully interpretable, training-free rule-based system grounded in 67 linguistic features spanning morphology, syntax, vocabulary, and syllabic structure. Gemini 3 and GPT-5.5 achieve the highest exact-match agreement on fine-grained reason prediction (55.8% and 56.4% Category+Reason), while the rule-based system achieves the strongest exact-match performance at the Category level (65.9%). Together, BAREC-X and our benchmark establish a transparent foundation for interpretable Arabic readability explanation.
%U https://aclanthology.org/2026.aacl-long.156/
%P 3445-3472
Markdown (Informal)
[Benchmarking Fine-Grained Readability Explanations for Arabic](https://aclanthology.org/2026.aacl-long.156/) (Rabih et al., AACL-IJCNLP 2026)
ACL
- Nour Rabih, Khalid N. Elmadani, Hanada Taha, and Nizar Habash. 2026. Benchmarking Fine-Grained Readability Explanations for Arabic. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3445–3472, Hengqin, China. Association for Computational Linguistics.