In “Learning Word Embeddings from Glosses: A Multi-Loss Framework for Arabic Reverse Dictionary Tasks” (Proceedings of The Third Arabic Natural Language Processing Conference), Engy Ibrahim (Alexandria University) and others “address the task of reverse dictionary modeling in Arabic, where the goal is to retrieve a target word given its definition. The task comprises two subtasks: (1) generating embeddings for Arabic words based on Arabic glosses, and (2) a cross-lingual setting where the gloss is in English and the target embedding is for the corresponding Arabic word.”
