In a
classic cartoon, two psychiatrists are depicted in the first panel as approaching each other, each saying “Good
morning.”
In the
second panel, as they walk away from either other, each has a thought-balloon,
thinking: “I wonder what he meant
by that?”
It’s
supposed to be a joke on psychiatrists;
though indeed, anyone familiar with psychology and linguistics
recognizes that to be a quite valid, indeed deep, question.
~
So
this morning, we have a “Good morning” scandal. It has been reasonably widely (though not well) reported
in the mainstream world press, though so far not the American. E.g. a high-circulation popular
Parisian daily:
http://www.leparisien.fr/politique/israel-un-palestinien-arrete-apres-une-erreur-de-traduction-de-facebook-22-10-2017-7348206.php
La police israélienne a arrêté par erreur
la semaine dernière un Palestinien qui avait publié sur Facebook une photo de
lui accompagnée des mots «bonne journée ». Selon le quotidien israélien «Haaretz»,
pour une raison encore inexpliquée, le logiciel de traduction du réseau social
a converti cette phrase en «attaquez-les » en hébreu et «faites leur du mal »
en anglais.
The
BBC’s brief note on the matter is
sociolinguistically useless for readers who wish to figure things out for
themselves, since it shows neither the photo nor the Arabic:
http://www.bbc.com/news/world-middle-east-41714152
Israeli police arrested a Palestinian man
last week after a Facebook post he made saying "good morning" in
Arabic was mistranslated to read "attack them" in Hebrew, local media
have reported.
Police confirmed that the construction
worker was briefly held under suspicion of incitement but was released as soon
as the mistake was realised.
The post showed a photo of the worker
next to a bulldozer in the West Bank.
Such vehicles have been used to attack
Israelis in the past.
There is only one difference in lettering
between the colloquial Arabic phrase for "good morning to you all"
and "hurt them", pointed out The Times of Israel.
Actually
the Times of Israel doesn’t “point out” any such thing. It merely piggybacks off an earlier Haaretz report, and does
not give either the original Arabic nor the purported “one difference in
lettering” that would give the meaning “hurt them”. -- Indeed, to anyone familiar with printed Arabic, that
proviso is a cop-out anyway: The
(usual) unvoweled Arabic ductus is so low in redundancy that, “merely” by
changing “one letter”, you can
radically change the meaning of just about any short message. -- More anon.
Here
is the beginning of the original article in Haaretz:
https://www.haaretz.com/israel-news/1.818437
No Arabic-speaking police officer read
the post before arresting the man, who works at a construction site in a West
Bank settlement
That
detail provides an innocent explanation of why the fellow would take a selfie
with a bulldozer. He probably
drives the thing.
It
continues:
The Israel Police mistakenly arrested a
Palestinian worker last week because they relied on automatic translation
software to translate a post he wrote on his Facebook page. The Palestinian was
arrested after writing “good morning,” which was misinterpreted; no
Arabic-speaking police officer read the post before the man’s arrest.
That
bit about the dreadful consequences of not having an Arabist on your payroll is
a pleasing one (to Arabists). Compare this:
~
Back
to the psycholinguistic intricacies of “Good morning”.
The
linguistically naïve reader (here in the hands of linguistically naïve journalists,
so no help) will assume that “Good morning” in Arabic is completely straightforward, and that any translator --
human or machine -- that couldn’t translate that simple phrase, would be
utterly incompetent.
Now,
there is in fact a
straightforward, MSA phrase for ‘Good morning’ in Arabic,
صباح الخير
Phonetically (and we’ll stick to
phonetics from here on, since Microsoft often mangles Arabic script): ṣabāḥ al-xayr. Literally, ‘morning of goodness’ (syntactically, an idafa), to
which the usual Arabic morphosemantic rules apply.
But that is not what he wrote.
Einführung
in die vergleichende “Guten Morgen”-Morphosemantik
Of the
languages with which I am familiar, German comes closest to having a
word-for-word equivalent to “Good morning”: Guten Morgen. Yet even that is not linguistically
straightforward: the idiom
involves not guter, but guten -- grammatically an accusative. So, you understand the way the
phrase is used, but you probably have never reflected on its grammar, and perhaps
could not explain it if you did.
Further,
in parts of Germany (e.g. Schweinfurt, where I lived for a summer with a German
family), people don’t usually say Guten
Morgen anyway, they say Grüß Gott
-- a phrase I am fond of, and frequently use, but whose syntax is obscure; I suppose it’s a sort of iḍāfa.
In
French, a word-for-word equivalent to the greeting is not current. You don’t say “bon matin”, you say bonjour, literally ‘good day’, or (as in
the Le Parisien version) bonne journée
(the nuance of difference being untranslatable). (There does, however, exist in French an early-morning
greeting which we lack in
English: bon réveil. I’ve never
myself heard that, not being an early riser.)
In
Spanish, you must go even further afield, using a plural, buenos días, lit. ‘good days’.
As to
what such phrases are used for, there
are subtle distinctions. Thus, in
current American English, Good morning!
is simply a greeting -- a pure illocution, like hi or hello, with no
descriptive content. (Note: In British English, it can serve as a good-bye, even as a curt dismissal. Thus, the heavy father in Wodehouse's Pearls, Girls, and Monty Bodkin says "Good morning, Mr Bodkin" to the unwelcome suitor of his daughter, signaling that their interview is at an end.) Good night!, by contrast, cannot be used
as a greeting: it is a formula of
leave-taking -- a valediction. Good day!, the word-for-word equivalent of the French greeting bonjour, is no longer used in everyday
American English, though it survives as a breezy greeting in Australia -- the
iconic G’dye! Good
afternoon and Good evening can
still be used in the U.S., though they sound a bit formal. And at least the latter has also been
used (at one time, primarily perhaps in England) as a valediction.
As
mentioned, French doesn’t normally use “Bon matin”, at least not the variety of
French I’m familiar with. Bonne matinée might be used but, like bonne journée, my hunch is that it is
not so much a pure greeting as an optative,
relating to what might follow the encounter, like the trademarked American “Have
a nice day!” (I first heard that
one in Berkeley, California, decades ago;
since then it has spread like kudzu throughout the English-speaking lands.)
As for
Guten Tag, there’s that accusative, a
fossilized echo of its likely origin in such a phrase as “Ich wünsche
Ihnen einen guten Tag”, which is an optative. Do contemporary speakers feel this influence, in which case Guten Morgen is not after all an exact
equivalent of Good morning ? Hard to say.
In Modern Standard Arabic (and in most dialects), ṣabāḥ al-xayr corresponds pretty closely to “Good morning”, but things get colorful from there. Although you may respond in kind, usually (in line with the Koranic injunction, “When you are greeted, respond with the same greeting, or a better one”) you say something like ṣabāḥ al-nūr (‘morning of light’), or ṣabāḥ al-full or ṣabāḥ al-yāsimīn (both meaning ‘morning of jasmine’).
In the
dialects, things get wilder. Thus,
in Yemeni, a very characteristic morning greeting is kaṣbaḥtu,
to which the standard reply is baḥḥakum Allah bi-xeir (sic; the initial
ṣ has been dropped). And --
save the mark! -- there even exist untranslatable Yemeni expressions involving
the ṣ-b-ḥ root, which my Sanaani teacher explained as “a morning warning” --
thus potentially, in line with the interpretation of the Israeli police, a threat.
However, what the Palestinian really
wrote was:
يصبحهم
And in
view of what has gone before, you will no longer be inclined to jump to
conclusions that you understand all the nuances of that.
First,
we cannot without further
analysis give a phonetic
transcription of that, since it is a script sans diacritics. Morphologically, it’s a
third-person-masculine-singular transitive verb, with
third-person-masculine-plural direct object. The verb could be either Form I, II, or IV, all of
which have transitive uses.
So, how
did the translator the Israelis relied on
parse it?
The “automatic
translation software” is not identified, though its digital facepage is
doubtless blushing magenta.
I just tried it in Google Translate, and (taking the verb as form-IV) it
renders the Arabic phrase as “become
them”, so the mystery remains.
The
most straightforward assumption is that we are here dealing simply with the MSA
verb, form II (syntactically transitive, semantically delocutive), in which case the expression means “He says ‘Good morning’ to
them.”
That
is actually a somewhat strange thing to say. You’d expect more like “(I say) “Good morning” (to
you(-all))”. “He” makes
sense as referring to the fellow in the photograph; but who are ‘they’? The object-pronoun is definitely third
person (and thus, the BBC rendering as "good morning to you all" is
incorrect).
Whether
there is some further nuance in Palestinian, I do not know offhand. And as for a translation as “attack”,
with the information we have been given so far, it is inexplicable.
~
For another cautionary-tale about the perils of interpreting brief swatches of Arabic, try this:
http://worldofdrjustice.blogspot.com/2017/10/rorschach-philology.html
[Appendix] The article that prompted this
post was of the “Mistranslation
Howlers” motif; my reply was
partly out of annoyance with that complacent genre, which often finds
linguistic illiterates gloating over
anecdotes that turn out either to
be more gray-area than appears at first blush (such as the one above), or to be
urban legends. A likely example of
the latter is the hardy perennial about the Chevy Nova being marketed among
hispanophones under that name.
Laymen observe triumphally that, in Spanish, “No va” means ‘it doesn’t
go’; epic fail !! Actually, nova is a perfectly normal word of
Spanish astronomical terminology (https://es.wikipedia.org/wiki/Nova ); it is neither spelled like “no va”
(being one word, not two), nor pronounced the same (being forestressed).
The “Good
morning” article was at the expense of the Israeli counter-terrorist
authorities. There seems indeed to
have been a misstep in the software, though one more subtle than most accounts
suggested. The article first
appeared in Haaretz, itself an Israeli publication; its appearance may have been unmotivated other than by the
universal Schadenfreude over (alleged) mistranslations (some of which pass into
legend, like “Ich bin ein Berliner”, whether or not the interpretation is
linguistically well-founded), or there might have been a little Israel-internal
scalpel to grind -- nescio. But I would like to add a further
observation, of respect for those who must work the intricate and dangerous CT
mission.
Notoriously,
terrorists (and criminals) use cover-terms. The IC took a beating over a now-celebrated message
intercepted (we are told) from Afghanistan on September 10, 2001: “The match
begins tomorrow.” The post-hoc Besserwisser wagged their fingers and
berated the Agency for not getting that message translated until September 12
(which, given the exotic language of the message, and the skeleton crew left in
CT after the relentless whittling-down of the Clinton and Bush years, is only
too understandable.) More to the
point than the timeliness, is the interpretation: Was it really obvious, or should it
have been obvious, that this particular message, among a mass of similar
things, foretold the 9/11 attacks?
I personally have no idea (though a later news story did aver that the
message was eventually determined to have referred, in fact, to an upcoming
soccer-match).
Now,
as mentioned, the message in the current case is rather cryptic:
يصبحهم
It is
not (according to a Comment to this post, below, from a colleague with
significant language experience in Palestine) a particularly usual way to wish
someone (or some third parties, in this case) top o’ the morning. Might it have been an allusion, a wink-wink, or at least
contextually suggested one, to the Israel CT team? In other words, we need not assume that the team just
blindly-blandly took the Hebrew mistranslation as the final word on the subject. They might (given their special knowledge of the folkways of their target-set)
actually have known what they were doing.
But
surely (objects the straw-man, less savvy than yourself), something as simple
as a verb meaning ‘to offer a morning greeting’ could not be re-interpreted to mean anything nefarious. Yet as present company has
learned, language has more tricks up its sleeve than most folks give it credit for.
Consider the following famous phrase of German, which contains a verb
meaning ‘greet’:
Und täglich grüßt das Murmeltier
Now, I
know German reasonably well; but
when I first encountered that expression, my hair did stand on end (like quills
upon ye fretful Murmeltier).
It has the d’outre-tombe knell
of those mystery radio-phrases that Cocteau (with a wry nod to maquisard comsec during WWII) stuck into
his 1950 movie Orphée ("L'oiseau chante avec ses doigts"). What could it possibly mean? Were it my day-job to interpret such
things, if it came up in traffic
my first instinct would be to shut down all our embassies instanter. Yet it is
merely the German version of the movie-title “Groundhog Day”.
Inghimasi Murmeltier ... greets you |
.
What about the greeting "Hi" in American English? Or the word "hello"?
ReplyDeleteThe translation software must have been using a very violent text as its language model, where the closest match to the cheerful يصبحهم (good morning) was اذبحهم (slaughter them). The more formal greeting is spelled out in full on this FB page: https://ar-ar.facebook.com/AllahYsabehkomBelkhair/
ReplyDeleteIt's a form II verb: aḷḷa yṣabbiḥkum bi-l-xēr. Its evening counterpart is aḷḷa ymassīkum bi-l-xēr. These are MSA expressions, and there's probably an explanation for the use of the 3rd person -hum in the short form yṣabbiḥhum in colloquial Palestinian.
Thank you! Extremely helpful. -- For non-Arabists, the proposed violent etymon is, typographically, ludicrously far away from what was actually written.
DeleteYour suggestion that, nevertheless, that substitute might have cropped up in a special MT algorithm based on a very limited, very violent Vorlage, such as IDF is no doubt confronted with daily, is quite intriguing. I wonder if anyone has done a lexicon of Arabic based precisely and exclusively on terrorist texts.
It is not ludicrously far away, one is
ReplyDelete"isba7hum" (Say good morning to them (misspelled))
the other is
"idhba7hum" (Slaughter them)
and the innocent "isba7hum" can receive a dot on top of the s (the letter sad) to become the letter "dhad".. a different dh than in the murderous idhba7hum (dhal, not dhad), but they are often mixed up, pronounced the same by many, and it is quite possible the author intended to say that or to do a murderous word play.
Your observations are correct and well-taken. To the human eye, handwritten or faxed Arabic, where dots get lost or are added as static, a ṣād and a ḍād are easily confused. Apart from that, the pronunciation of ḍād does indeed conflate in the dialects -- though with ẓā, not with dhāl, so far as I know.
DeleteHowever, the script in question was neither handwritten nor faxed; moreover, context was a discussion of Machine Translation (allegedly used by the Israelis in this case -- “web-based”, according to the Times of Jerusalem). And a machine, blindly following its algorithms, cannot ‘look over its shoulder’ at neighboring words that, in human hands and in political context, might suggest murderous word-play. So the purported MT goof is still unexplained.
Thank you for the illuminating discussion of the phrase, which was poorly explained by every newspaper article I found.
DeleteAs a computational linguist, I thought I would just briefly comment about the machine looking at neighbouring words. Facebook now uses neural networks for its machine translation. While I don't know the details of their algorithm, it's quite possible that they use character-by-character processing in order to deal with misspellings and inflections it hasn't seen before. That might not be the cause of the mis-translation in this case, but it is possible. A neural network makes weird mistakes sometimes, especially when given input that's quite different from anything it was trained on.
Another thing to note is that they probably translated to Hebrew. Translating the phrase to Hebrew, then English yields "make them" which is far cry from "become them", so maybe that translation has the negative connotation that caused the situation.
ReplyDeleteExcellent post and comments.
ReplyDeleteThis story was quoted recently (3 years after the story appeared) at a Machine Translation conference as an example of poor MT. But the story and the various newspaper articles (when I did some Googling) made no sense to me. So thank you for explaining it properly and so entertainingly too. It struck me that if the person in question was using an English keyboard to type Arabic phonetically then the letters "s" and "d" could easily be typed in error.
ReplyDeleteThe arabic chat alphabet
ReplyDeletehttps://en.m.wikipedia.org/wiki/Arabic_chat_alphabet