[numpad0]

I don't think it did for Korean, though I need input from speakers to be sure. From my experience, Korean MT routinely stops halfway through inputs and dumps nonsensical phonetic transcripts, likely from failing to identify words. I suspect they were just being complaisant to American influence in postwar years. Computers failing to even isolate and match words in this day and age is not a sign of an excellent working script.

[yorwba]

Translation needs phonetic transcription to handle proper names. If there are words that may or may not be proper names depending on context, machine translation will guess the context wrong at least some of the time and phonetically transcribe what should be translated, or translate names that should be transcribed.

The problem also can also happen when translating from English, if you think about all the surnames that are occupations, or names like "bill" or "lily." Capitalization usually helps disambiguate, but there's title case and all caps and people who never capitalize anything...

[numpad0]

It's not just proper nouns. Korean MT seem to routinely "de-synchronize" into wonbonhangugeotegseuteububun mid-sentence and sometimes comes back in sync, sometime stays out of sync until the end of the sentence. it happens way more often than average with the Korean language.

[yorwba]

Do you have an example input where that happens?