Translating the Untranslatable: An Operationalizable Ontology for Untranslatability

This article presents a structured ontology for untranslatability in machine translation (MT), addressing the limitations of current MT systems in handling cases where meaning cannot be preserved across languages. It introduces a multilingual dataset of untranslatable sentences paired with compensation strategies, enabling controlled analysis of translation behavior. The findings indicate that translation quality varies based on the strategy employed, highlighting the importance of context in improving MT outputs, which is crucial for practitioners developing more effective translation systems.