Hungarian machine translation post-editing

Two levels, chosen deliberately

• Full post-editing — the result reads as though it was translated by a human: accurate, fluent, consistent terminology, correct register. Use it for anything a customer sees.
• Light post-editing — errors that change meaning are fixed, the rest is left alone. The text is correct and usable but not polished. Use it for internal content, high volumes, or material with a short shelf life.

Where MTPE works, and where it does not

Machine output is a reasonable starting point for repetitive product content, documentation and support material. It is a poor starting point for marketing copy, anything with wordplay, and crypto content where a mistranslated term can mislead a user about money. For those, translating from scratch is usually faster than repairing the output, and I will tell you when that is the case rather than quietly charging post-editing rates for what is really a retranslation.

Hungarian is a hard case for machine translation

Hungarian is agglutinative, with case endings that change depending on the word they attach to and on vowel harmony. Machine output tends to be grammatically plausible and subtly wrong: the right word in the wrong case, a placeholder that breaks the sentence it sits in, a formal register where the product speaks informally. These are exactly the errors a reader notices and a spellchecker does not.

Experience behind it

I perform MTPE on a regular basis across several accounts, including Microsoft Xbox and Microsoft Learn content through Jonckers, alongside translation and review work on the same material.