System Submission
Participants submit systems that predict both broad and specific genre labels for Arabic text segments using the provided hierarchical taxonomy and English genre definitions.
NLP/CORE Unranked
AraGenre 2026
ArabicNLP2026-AraGenre2026
AraGenre 2026 is a shared task focused on hierarchical Arabic genre classification, evaluating systems' ability to identify both broad and specific communicative genres in Arabic texts under low-resource conditions. It emphasizes robust semantic understanding over lexical memorization by using a definition-guided framework with synthetic training data and a hidden evaluation set of noisy, naturally occurring texts across dialects and writing styles.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Hierarchical Genre Classification
Broad Genre: InformativeBroad Genre: CreativeBroad Genre: InteractiveBroad Genre: LearningBroad Genre: LegalBroad Genre: ReligiousSpecific Genre: Analytical ReportsSpecific Genre: Educational ExplanationsSpecific Genre: Diplomatic CommunicationSpecific Genre: Motivational WritingSpecific Genre: Forum DiscussionSpecific Genre: Religious Commentary
Policies worth checking twice
- Systems are evaluated using Hierarchical Macro F1 as the official ranking metric.
- Training and development data are synthetic and limited to simulate low-resource conditions.
- The evaluation benchmark contains naturally occurring, noisy Arabic texts with previously unseen genre-definition combinations.
- Texts span Modern Standard Arabic, Classical Arabic, and multiple dialects, including diacritised and undiacritised forms.
- Systems must assign both a broad genre and a specific genre label for each input text segment.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026