This is the most 2023 answer possible but I wonder if LLM translation will help handle this sort of situation better.
An LLM first translates to tokens in more of a "concept space", then translates to other languages. So it would translate "zip file" to something completely different than "zip code", and translate to UK English correctly, or at least better I would hope.
They may not solve all problems, but maybe this one?
An LLM first translates to tokens in more of a "concept space", then translates to other languages. So it would translate "zip file" to something completely different than "zip code", and translate to UK English correctly, or at least better I would hope.
They may not solve all problems, but maybe this one?