Mark Liberman has a post at the Log quoting Tal Linzen reporting that Google Translate renders Hebrew “Please return to me” as “Please me like an alien creature”:
The first word אנא [‘ana] means ‘please’ (though only in the request sense) and the last word אלי [e’laj] means ‘to me’. The source of the mistranslation is the second word חיזרי, which can be either [xiz’ri] ‘return (imperative, singular, female)’ or [xajza’ri] ‘extraterrestrial (adjective)’.The spelling for the ‘return’ sense is typically חזרי, but it’s not that unusual to spell it with the vowel חיזרי. A more common spelling for the second form would be חייזרי, doubling the י to indicate that it’s used as the glide [j] rather than the vowel [i]. Regardless of the spelling, though, it’s surprising that ‘extraterrestrial (adjective)’ is more frequent than ‘return (imperative)’.
Hilarious as this is, the reason I’m posting about it here is an interesting disagreement in the comment thread. David L. Gold writes:
In non-formal latter-day Israeli Hebrew, most of the historically imperative forms are not used. Rather, future-tense forms are used. […]
For chazar, the historically imperative forms (masculine singular chazor, feminine singular chizri, and plural chizru) are not used (remember, I am speaking here about non-formal latter-day Israeli Hebrew; those three forms ARE used in formal Israeli Hebrew).
Rather, in non-formal latter-day Israeli Hebrew, the imperative forms are the future-tense forms:
masculine singular tachzor! (literally, ‘you will return!, you shall return!’)
feminine singular tachzeri (literally, ‘you will return!, you shall return!’)
plural tachzeru! (literally, ‘you will return!, you shall return!’)
Ran Ari-Gur responds:
That’s true in spoken Hebrew, but in written Hebrew imperative forms aren’t nearly so uncommon; and one would have expected Google’s bilingual corpora to bias toward the latter. (Googling “חיזרי” and looking through the first few pages, by the way, I find that almost all hits are instances of /xiz’ri/; very few are /xajza’ri/.)
I’m guessing that Yoav Goldberg has it right; if this translation came from a dictionary rather than from corpus analysis, then the lemma form /xajza’ri/ would have had an advantage over the non-lemma form /xiz’ri/ (lemma = /xa’zar/).
Both reiterate their points; Gold says:
In latter-day Israeli Hebrew, the controlling factor in the choice between (1) historically imperative forms (such as !חזור) and future-tense-forms-used-as-imperatives (such as !תחזור) is not the means (written or oral) by which the utterance is conveyed.(*)
Rather, the controlling factor is the location of the utterance (be it written or spoken) on the continuum of (in)formality.
It’s true that, all else being equal, the true imperatives are more formal and the future-tense-forms-as-imperatives are more informal — and I don’t think I implied otherwise. But I stand by what I wrote. In general, true imperatives occur much frequently in writing than in speech. (In part, this is because people tend to write more formally than they speak; though I’m not sure if that’s the whole story. To some extent I think the converse may also be true, that people sometimes affect formality in speech by borrowing elements of a written style, and vice versa.)
I’m curious to know what my readers have to say about it.