When It Rains Ropes
François Yvon, a research director specialising in automatic translation at France’s National Centre for Scientific Research CNRS, says idioms are a classic test. A good system will understand that ropes are not literally falling from the sky. But language is full of subtler traps.
When a machine translator comes across the sentence ‘translate into Polish’, it is quite capable of deciding it must be all about shoe cream.
Context is crucial. If a system has not grasped that the subject is a football match, it can easily mistranslate “taking a corner”.
A Trillion Words a Month
The scale of machine translation today is staggering. A few key figures:
Google says people translate one trillion words a month with Google Translate, not counting DeepL, Microsoft, Claude, ChatGPT and Gemini.

















