Machine translation often lacks understanding of context.
Innocent words in one language -given a mistaken context- can be translated into bad words.
Sensitivities in one language are not those in other languages.
e.g. the word for a female dog in english is one that's frowned upon, but what if it's not used as such in the source language ?
e.g. in my native language the neutral name for somebody with black skin is not frowned upon. It's not abad word in itself (describing races is more frowned upon than e.g. in the US). Translate it to English and half the world calls you a racist.
So: no thanks: no machine translation of my content, let alone authorizing advertising on it where the advertising agency overreacts to "bad" content.
Context is hard: it requires the translator to understand the text itself.
- The kind made at Intel ?
- The kind you car has when it get's stuck by a pebble ?
- The kind you buy in a air-sealed bag in a gas station?
- French Fries ? (as in "fish and chips").
Rest assured that in many languages the translation for chips isn't all into the same word for every kind of Chip there is in English.
E.g. in French "la ferme" can mean
- shut up! (a rather impolite and urgent way to say it) - a better translation would probably be STFU.
- the farm
I've seen professional (human made) translations where they got it wrong. Hilarious in itself - till you start to use bad words in another language.
Context can also be regional.
Don't forget that in many languages bad words have an origin in innocent words e.g. a text describing how to bolt something together could easily be translated into something using "screwing" and "nuts".
[edited by: swa66 at 2:28 pm (utc) on Feb 25, 2013]
Agree 100% with swa66..
Oops I elaborated after Leosghost posted that ... sorry had not noticed the reply or I would not have done that.
No problem :) I still agree 100% with you..even when what is written on the bags of potato based snacks in France is "crisps"..French people still all pronounce the word as "chips"..or more often as "ships" or "sheeps" :)..
In French the word "Gueule" for example often gets translated to "mouth" in English..It does mean mouth , but only as in mouth of an animal ..to refer to the mouth of a person as a "gueule" is to insult them..as in "ta gueule" ( does not mean "your mouth" )..means STFU..
Although there is apparently a French restaurant in Japan which goes by the name of "a ta gueule" :)
There is also "gueuler" [en.wiktionary.org...] meaning to yell or shout and the derivatives of it ..all contextual..and all totally beyond automated ( and even as swa66 said, some so called "professional" human ) translation ..
|Someone in a recent thread said that they did not include any google translate sites in their list of authorised sites for Adsense for "obvious reasons". |
i said "good reasons" so this wasn't me then..
|Why wouldn't you authorise these Google sites? |
the context of the thread was the use of a framebuster script.
since the google serps translate link opens your webpage in a frame (which btw is illegal just as their new image search), the framebuster script gets triggered and opens your homepage as parent window. but sometimes the adsense in the framed window seems to be loaded before the framebuster script is triggered. as i see it, if you don't want to be associated with this worthless page view in milliseconds, you will have to de-authorize google translate so you won't get credited. other reasons to do that seem indeed possible.
|Innocent words in one language -given a mistaken context- can be translated into bad words. |
Oh, ###. You mean naaraajiiruluk doesn't mean "My esteemed colleague from Kuujjuaq" after all? I'm in trouble now...
But seriously: Doesn't even have to be translation. A machine can do just fine without ever leaving its home language. Uhm, no, google, I don't need any ads for exterminators, thanks all the same. Conversely it only takes a moment of absent-mindedness for a human to achieve something of machine-translation caliber. As when the Italian version of my page about typing Greek spoke of a chiave (=clef) rather than a tasto (=touche, I guess). Oops.
|Although there is apparently a French restaurant in Japan which goes by the name of "a ta gueule" |
What-- you thought Japan only made war on the English language?
I'm guessing there's an etymological connection to "gullet" in there somewhere.
I'm sure you realize that calling something illegal doesn't make it illegal. It's the rare law that is beyond interpretation. ("But he took my parking space! Any jury in the world would agree it was justifiable homicide.")
Thank you everyone. I couldn't remember the original post but I think i must have mis-quoted moTi.
Shall have another look at my authorised list, having (very) naively thought anything Google in the list should be ok.
well, in case you have a framebuster script installed, i can't tell you if the said approach is necessary - maybe even google can't.
for me, only authorizing my own websites is just precaution. ymmv
In my case, most of my AdSense sites are targeted to US states. No point to messing with translations.
I have no problem with translation sites and I have a whole load of them authorised, mostly Google.
Whether they are good translations I don't know, I can only go on the reverse of translating other languages into English which I find more than acceptable for my purposes. As long as I get the general thrust of things I'm happy.
I don't have a problem with authorizing Translate for AdSense, the ads seem targeted to the user the same as normal.
I lose control over who is accessing the site via G Translate, I don't receive meaningful data for those people. IP banning won't work on them. It just seems like an exploit waiting to happen. So I prefer to close the door and keep it shut.
|I lose control over who is accessing the site via G Translate, |
You can't do a simple CIDR block but you can still get the info from the X-Forwarded-For header. G### unlike some other search engines at least sends one. I get them from Preview, Wireless Transcoder and
:: shuffling papers ::
yup, here's one from Translate:
User-Agent: Mozilla/5.0 (Windows NT 5.1; rv:19.0) Gecko/20100101 Firefox/19.0,gzip(gfe)
Via: 1.0 translate.google.com TWSFE/0.9
X-Forwarded-For: 83.31.114.nnn, 83.31.114.nnn
(Further detour tells me it's one of those beastly Polish ranges that's a random mix of humans and robots.)