| Attachment | Size |
|---|---|
 poster4.jpg | 84.52 KB |
KnowledgeGraphEval 2026 Shared Task at ArabicNLP 2026
Dear All,
We invite researchers and practitioners to participate in KnowledgeGraphEval 2026, a shared task at ArabicNLP 2026 on Arabic knowledge graph construction and domain adaptation.The shared task focuses on building structured knowledge representations that support downstream applications such as information retrieval, semantic search, question answering, and modern NLP systems, including Retrieval-Augmented Generation (RAG) and Large Language Models (LLMs).
To simplify the construction of knowledge graphs, we decompose the problem into two core tasks that are well established within the NLP community: named entity recognition and semantic relation extraction.KnowledgeGraphEval 2026 includes two subtasks:
• Cross-Domain Named Entity Recognition (AdaptNER): Adapting NER models across 10 different domains.
• Relation Extraction (RE): Identify and classify semantic relations between entity pairs.
The shared task emphasizes robust evaluation under both in-domain and out-of-domain settings across multiple domains.
We also introduce two new resources: Konooz, a large-scale multi-domain and multi-dialect Arabic NER dataset, and WojoodRelations, a large-scale Arabic relation extraction dataset.
Datasets, baselines, and evaluation scripts are available on the website.
Website: https://sina.birzeit.edu/KnowledgeGraphEval
Registration: https://forms.gle/xGuSr76W6jnxV7At9
Important Dates
- May 16, 2026: Release of the shared task website and registration form.
- June 5, 2026: Release of training/development data and evaluation scripts.
- July 20, 2026: Registration Deadline and Blind Test Data Release.
- July 30, 2026: Official results and rankings are released.
- August 22, 2026: Deadline for Camera-ready for Participant System Description Papers.
- October 24-29, 2026: ArabicNLP 2026 / EMNLP 2026 - Shared task overview and participant systems presented in Budapest, Hungary.
We encourage researchers and students working on NLP, information extraction, LLMs, domain adaptation, and knowledge graphs to participate.
Best Regards,
The KnowledgeGraphEval 2026 Organizing Team