mersenneforum.org Interesting table-lookup bug in Google Translate
 Register FAQ Search Today's Posts Mark Forums Read

 2010-12-08, 02:30 #1 ewmayer ∂2ω=0     Sep 2002 República de California 267208 Posts Interesting table-lookup bug in Google Translate Was using GT to try to get an English translation of a legal maxim just now. The aphorism in question is "Fraus et dolus nemini patrocianari debent." (Fraud and deceit should excuse no man - found the correct rendition later at a legalese-oriented site) Google Translate returned this: "And justice never agree together." Which is both wildly off and drops the obvious leading word "Fraus" (fraud) completely. That had me baffled until I found the following similar-sounding maxim at the same legal-lingo site as above: "Fraus et jus numquam cohabitant." (Fraud and justice never agree together.) It seems GT is doing some kind of table-lookup-by-leading-word either to save time or when its internal algos fail to converge on a translation of the input phrase; and when it does this lookup-method it erroneously deletes the leading word used for the lookup. Lets try another one, "Communis error no facit just." (A common error does not make law.) Google mangles this one as "I do the common error of hidden truths." No idea what the heck kind of algo it used there. :P
2010-12-08, 20:49   #2
ewmayer
2ω=0

Sep 2002
República de California

24×733 Posts

Let's try things going the other way, and see how GT does at translating the Lyrics to Jimi Hendrixs Purple Haze first into Latin, then back into English:

Original (English):
Quote:
 Purple haze all in my brain Lately things just don't seem the same Actin' funny, but I don't know why 'Scuse me while I kiss the sky Purple haze all around Don't know if I'm comin' up or down Am I happy or in misery? Whatever it is that girl put a spell on me Purple haze all in my eyes Don't know if it's day or night You got me blowin', blowin' my mind Is it tomorrow, or just the end of time?
The Latin Gospel "Purple Nebula" according to Google:
Quote:
 Purple obducto MEUM CEREBRUM omnes Iusta et e non videntur idem Dictyostelium discoideum actin 'funny, sed nescio quid 'Mihi donec Scuse caelum osculum Purple nebula circum Nesciant I'm Comin 'sursum deorsumque Sum felix aut miser? Quodcumque puella dedit mihi carmine Purple nebula oculos omnia Nescio si suus dies noctesque Tu possedi blowin 'blowin' animo Cras est, seu in finem saeculi?
"Dictyostelium discoideum"? If I translate just that back to English using GT it gives "Translated Amino Acid Sequence", neither of which makes *any* sense w.r.to the original text, and which short phrases differ wildly from each other, to boot. OK, dude, whatever you say ... but the rest was perhaps a bit unfair in terms of abbreviations and contractions ... Let's render all the apostrophized bits in the original in plain English, e.g. "'scuse me" --> "excuse me" ... that leads to this, which magically has dropped the gene-splicing-experiment-gone-awry:
Quote:
 Purple obducto MEUM CEREBRUM omnes Iusta et e non videntur idem Tum funny, sed nescio quid Dum dog caelum osculum Purple nebula circum Nesciant venio sursum deorsumque Sum felix aut miser? Quidquid id est, puella dedit mihi carmine Purple nebula oculos omnia Nescio si dies noctesque Me habetis flante flante animo Cras est, seu in finem saeculi?
And in a classic feedback loop, we translate that back into English to see if the there-and-back-again voyage was at all lossy:
Quote:
 Purple haze all My brain A reasonable one, and vice do not seem to the same Then the funny, but I do not know what While not broke, the kiss of heaven Purple a cloud around the Do not know how I come up and downwards the I am happy or to be unhappy? Whatever it is, the damsel gave me by a poem A cloud the eyes of all things Purple I do not know if the day and night Me, you have the breath of the breath of his mind Tomorrow is a, or in the end of the world?
Ha, ha ... I could do several more cycles of this, but I'm afraid I might pull a ligament laughing.

 2010-12-08, 21:52 #3 xilman Bamboozled!     "𒉺𒌌𒇷𒆷𒀭" May 2003 Down not across 2×32×17×37 Posts A favourite saying of mine, which I use quite often, because it is so apt is "blind idiot" (as in "it's a case of blind idiot") . See if you can work out what it was in the original English before translation back into English from intermediate Russian. Using Google is cheating. Paul Last fiddled with by xilman on 2010-12-08 at 21:53 Reason: Fix tyop and, punctuation
 2010-12-08, 23:15 #4 Batalov     "Serge" Mar 2008 Phi(4,2^7658614+1)/2 23·1,229 Posts Dicty is a popular amoeba for biological research. I am not sure Jimi would be happy with that whole translation! There was some site that google-translated 50 times ad absurdum. http://translationparty.com/ ? http://www.conveythis.com/translation.php ? or some other... don't remember which one was good. Last fiddled with by Batalov on 2010-12-08 at 23:28
2010-12-09, 00:24   #5
ewmayer
2ω=0

Sep 2002
República de California

1172810 Posts

Quote:
 Originally Posted by Batalov Dicty is a popular amoeba for biological research. I am not sure Jimi would be happy with that whole translation!
OK, so we know the species, but "where in blazes did that come from?" still applies. Does that mean GT suffers from amoebiasis?

Not sure what Jimi might have thought of the GT-squared alternate rendition ... you must admit it has a kind of hallucinogenic mellifluosity to it, somewhat like a drunk Ukrainian dockworker bellowing the lyrics to the song as best he can remember them after one hearing in a vodka-fueled karaoke session with his buddies.

"друзья, преступников, соотечественников: Do not know how I come up and downwards the!!!"

 2010-12-09, 00:55 #6 Batalov     "Serge" Mar 2008 Phi(4,2^7658614+1)/2 23·1,229 Posts ...If you translate it a few more times with conveythis, it will probably converge to the 'Bezerker' from the original Clerks.
2010-12-09, 01:32   #7
KingKurly

Sep 2010
Annapolis, MD, USA

33×7 Posts

Quote:
 Originally Posted by Batalov ...If you translate it a few more times with conveythis, it will probably converge to the 'Bezerker' from the original Clerks.
Congratulations, I literally laughed out loud.

2010-12-09, 01:49   #8
science_man_88

"Forget I exist"
Jul 2009
Dumbassville

838410 Posts

Quote:
 Originally Posted by ewmayer Was using GT to try to get an English translation of a legal maxim just now. The aphorism in question is "Fraus et dolus nemini patrocianari debent." (Fraud and deceit should excuse no man - found the correct rendition later at a legalese-oriented site) Google Translate returned this: "And justice never agree together." Which is both wildly off and drops the obvious leading word "Fraus" (fraud) completely. That had me baffled until I found the following similar-sounding maxim at the same legal-lingo site as above: "Fraus et jus numquam cohabitant." (Fraud and justice never agree together.) It seems GT is doing some kind of table-lookup-by-leading-word either to save time or when its internal algos fail to converge on a translation of the input phrase; and when it does this lookup-method it erroneously deletes the leading word used for the lookup. Let`s try another one, "Communis error no facit just." (A common error does not make law.) Google mangles this one as "I do the common error of hidden truths." No idea what the heck kind of algo it used there. :P
Not sure what's happening lol I can look through Google translates html with Google developer tools but I'm not great at decoding it lol.

 2010-12-09, 01:58 #9 ewmayer ∂2ω=0     Sep 2002 República de California 24×733 Posts "Whatever it is, the damsel gave me by a poem!!" Urrkkk Woohoo!
2011-02-06, 02:01   #10
Flatlander
I quite division it

"Chris"
Feb 2005
England

40358 Posts

Quote:
 Originally Posted by xilman A favourite saying of mine, which I use quite often, because it is so apt is "blind idiot" (as in "it's a case of blind idiot") . See if you can work out what it was in the original English before translation back into English from intermediate Russian. Using Google is cheating. Paul
I remember that one.

I also remember watching a BBC computer programme about the difficulties of computer speech recognition:
"Recognise speech" or "Wreck a nice beach"?

2011-02-06, 19:41   #11
xilman
Bamboozled!

"𒉺𒌌𒇷𒆷𒀭"
May 2003
Down not across

2×32×17×37 Posts

Quote:
 Originally Posted by Flatlander I remember that one. I also remember watching a BBC computer programme about the difficulties of computer speech recognition: "Recognise speech" or "Wreck a nice beach"?
Computers? Even native English speakers have trouble. There are several sites which record Mondegreens.

Some of my favourites include:
Is this the real life? Is this just Battersea?

The girl with colitis goes by.

Did you exchange, a walk-on part in the war for a negro in a cage?

Four candles.
(and many more from the same sketch)
And, of course,
Scuse me while I kiss this guy
which is pretty much where we came in.

Paul

 Similar Threads Thread Thread Starter Forum Replies Last Post ewmayer Programming 6 2012-11-12 00:25 Primeinator PrimeNet 2 2009-07-21 23:19 Joshua2 Software 5 2009-02-23 22:59 cwilliamsprime Teams 1 2007-02-23 20:55 ewmayer Soap Box 1 2005-08-09 14:13

All times are UTC. The time now is 14:16.

Sat May 28 14:16:01 UTC 2022 up 44 days, 12:17, 0 users, load averages: 0.62, 1.16, 1.30