Why does machine translation of legal contracts often fail to produce a coherent target text, even if each sentence is individually translated accurately?

💡 Explanation

Machine translation struggles with coherence because discourse relations (e.g., cause-effect, contrast) between sentences are often implicit and depend on legal context, and those discourse relations are poorly transferred. Therefore, the resulting text lacks overall coherence rather than showing errors in sentence-level translation, word knowledge or computational load.

• Why does mapping the abstract 'argument' domain onto the 'war' conceptual frame often fail to achieve persuasion within a target audience?
• If an experienced proofreader reads a document with inconsistent terminology, which consequence most likely follows regarding their eye movements?
• Why does syntactic regularization during creole formation frequently result in SVO (Subject-Verb-Object) word order, even when the substrate languages exhibit different dominant word orders?
• A cochlear implant processes sound to bypass damaged sensory cells. Which consequence follows if the implant's signal processing creates excessive spectral splatter?
• If a computational linguist modifies a statistical parser to favor shorter dependency lengths, which consequence follows?
• In a political debate, which communication strategy primarily leverages pathos to sway undecided voters?