๐Ÿ“ข We are also co-hosting ImageEval 2026, a shared task at ArabicNLP 2026, co-located with EMNLP.

The OASIS Dataset

The shared task is built on OASIS: images paired with spoken and written questions and open-ended answers, collected across the MENA region.

Open on Hugging Face

Whatโ€™s inside

OASIS pairs each image with a question and an open-ended answer, in both spoken and written form. The data was collected across 18 countries in the MENA region and spans everyday and cultural life. Questions come in English, Modern Standard Arabic, Egyptian Arabic, and Levantine Arabic, and every question is recorded as audio as well as text.

  • 18countries covered
  • 9topic categories
  • 31sub-categories
  • 4language varieties
Sample record image ยท audio ยท text
The National Museum of Qatar, its interlocking discs inspired by the desert rose
question_en
What is the name of the building shown in the image, and what inspired its design?
question_ar
ู…ุง ุงุณู… ุงู„ู…ุจู†ู‰ ุงู„ู…ูˆุถุญ ููŠ ุงู„ุตูˆุฑุฉุŒ ูˆู…ุง ุงู„ุฐูŠ ุฃู„ู‡ู… ุชุตู…ูŠู…ู‡ุŸ
question_audio
the same question, spoken
answer_en
The building is the National Museum of Qatar, and its design is inspired by desert rose formations.
answer_ar
ุงู„ู…ุจู†ู‰ ู‡ูˆ ู…ุชุญู ู‚ุทุฑ ุงู„ูˆุทู†ูŠุŒ ูˆุชุตู…ูŠู…ู‡ ู…ุณุชูˆุญู‰ ู…ู† ุชุดูƒูŠู„ุงุช ุงู„ูˆุฑูˆุฏ ุงู„ุตุญุฑุงูˆูŠุฉ.

An example record: the same question is asked as audio and as text, in English and Arabic varieties, and the target is a short open-ended answer. Photo: Msarg77, CC BY-SA 4.0.

Getting the data

The shared-task data is hosted on Hugging Face and is drawn from OASIS, or created with the same framework. A sample set is available now so you can preview the format; the training and development data follow on the schedule in the timeline.

License

The shared-task dataset is planned for release under the CC BY-NC-SA 4.0 license: free for non-commercial research use, with attribution and share-alike.

Citation

OASIS is described in the paper arXiv:2510.06371. If you use the dataset, please cite:

@article{alam2025everydaymmqa,
  title = {{OASIS}: A Multilingual and Multimodal Dataset for Culturally Grounded Spoken Visual QA},
  author = {Alam, Firoj and Shahroor, Ali Ezzat and Hasan, Md. Arid and Ali, Zien Sheikh and Bhatti, Hunzalah Hassan and Kmainasi, Mohamed Bayan and Chowdhury, Shammur Absar and Mousi, Basel and Dalvi, Fahim and Durrani, Nadir and Milic-Frayling, Natasa},
  journal = {arXiv preprint arXiv:2510.06371},
  year = {2025},
}

Ready to take part?

Registration is free, and the Slack community is where news lands first.

Register your team Join the Slack