Academic research project at MBZUAI building a multimodal benchmark of African cultural content (cuisine, attire, art, architecture, festivals). We have raw image–text data already collected and need native speakers to validate and refine it.
Task per sample
- Verify the image and text correctly represent the claimed country and category.
- Correct or rewrite the caption in your native language and English.
- Add brief cultural context where relevant.
- Flag mislabeled, low-quality, or duplicate samples.
Estimated 2–4 minutes per sample. Annotation done on our in-house platform (browser-based, no install). Short onboarding and written guidelines provided.
Languages needed (54)
Looking for native speakers in any of these. Multilingual annotators welcome.
- Arabic varieties: Tunisian (aeb), Egyptian (arz), Mesopotamian (acm), Najdi (ars), South Levantine (ajp), Levantine (apc), MSA (arb), Algerian (arq), Moroccan (ary)
- North & West Africa: Kabyle (kab), Twi (twi), Dyula (dyu), Fon (fon), Kabiyé (kbp), Mossi (mos), Bambara (bam), Fulah (ful), Akan (aka), Yoruba (yor), Nigerian Pidgin (pcm), Hausa (hau), Igbo (ibo), Umbundu (umb), Lingala (lin), Kimbundu (kmb)
- Central & East Africa: Kikongo (kon), Luba-Kasai (lua), Bemba (bem), Kinyarwanda (kin), Luganda (lug), Oromo (orm), Amharic (amh), Kirundi (run), West-Central Oromo (gaz), Kamba (kam), Kikuyu (kik), Nuer (nus), Chichewa/Nyanja (nya), Luo (luo), Somali (som), Swahili (swa)
- Horn & Indian Ocean: Tigrinya (tir), Malagasy (mlg)
- Southern Africa: Shona (sna), Xhosa (xho), Sesotho (sot), Zulu (zul), Tsonga (tso), Afrikaans (afr), Northern Sotho (nso), Tumbuka (tum), Tswana (tsn), IsiNdebele (nde)
- Lusophone Africa: Portuguese (por) — speakers from Angola, Mozambique, Cape Verde, etc., Emakhuwa (vmw)
Requirements
- Native or near-native fluency in at least one listed language.
- Strong cultural familiarity with the relevant region.
- Working English proficiency.
- Reliable internet; attention to detail.
Pay & volume
- Per-sample rate, weekly payouts via OpenTrain.
- Pilot: 50 samples per language. Full project: up to 1,000–2,000 samples per language over 6–8 weeks.
- Flexible hours.
To apply
Tell us: (1) language(s) and country/region, (2) brief note on cultural familiarity, (3) any prior annotation or linguistic work, (4) availability.