System Description Paper
Papers describing submitted systems for the shared task, following ArabicNLP guidelines, due by August 22, 2026.
NLP/CORE Unranked
ArabicNLP2026 - AlexandriaX 2026
AlexandriaX Shared Task @ ArabicNLP 2026
ArabicNLP2026 - AlexandriaX 2026 is a shared task focused on advancing context-aware dialectal Arabic machine translation and its evaluation. It addresses challenges such as regional linguistic variation, limited data, and sociolinguistic adaptation through three subtasks: dialogue translation, cross-dialect translation, and error detection with interpretable classification.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Shared Task Subtasks
Context-Aware English-to-Dialectal Arabic Dialogue TranslationCross-Dialect Arabic Machine TranslationMT Error Detection and Classification
Policies worth checking twice
- Submissions must be made via the official CodaBench competition pages.
- Constrained track for Subtask 1 restricts participants to using only the provided Alexandria dataset and models with ≤5B parameters.
- Participants may submit to one or more subtasks and tracks.
- Final rankings are based solely on official submissions before the deadline.
- Teams must not manually inspect or infer hidden gold labels during evaluation.
- Outputs must use provided IDs and follow specified formatting (JSON/CSV, UTF-8).
- Participants must run the released validation script before submission.
- A system description paper is required for participation.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026