EVALUATING NEURAL MACHINE TRANSLATION ACCURACY FOR SPECIALIZED ARCHITECTURAL TERMINOLOGY

Annotasiya

This article evaluates the accuracy and limitations of Neural Machine Translation (NMT) and Large Language Models in rendering specialized architectural terminology. The study identifies key semantic risks, including over-generalization and hallucination when processing culturally-bound architectural realia and historical vocabulary. The paper highlights the essential role of Machine Translation Post-Editing and proposes pedagogical strategies for integrating AI tools into translator training.