asbichi362/AfricaNLP2026_HCCLFN

HC²LFN: Hierarchical Contrastive Cross-Lingual Fusion Network for Machine Translation Detection in African Languages # AfricaNLP2026_HCCLFN HC²LFN: Hierarchical Contrastive Cross-Lingual Fusion Network for Machine Translation Detection in African Languages **Dataset Description** **Languages:** 20 african languages: NigerCongo (Swahili, Hausa, Yoruba, Zulu, Xhosa,Igbo, Shona, Kinyarwanda, Chichewa, Sesotho, Setswana, Wolof, Fon, Ewe, Akan, Twi) and AfroAsiatic (Amharic, Somali, Oromo, Tigrinya) families. **Sources:** FLORES+ devtest (1,012 sentences/language). **MT Systems:** Google Translate, NLLB-200, mBART-50, M2M-100, synthetic translations of GPT-4/Claude/Gemini. **Statistics:** 9645 samples of which 4000 were human and 5645 machine. 68% of this data was used for training, 12% validation and 20% for testing.