I look forward to their next efforts, American into English, Scouse/Manc/Geordie into English. And the hardest translation of all, Call Centre Monkey into anything anyone can understand.
Paris, cos she's good with her tongue as well
Google has announced the addition of ten new languages to its Google Translate feature, bringing the total of available tongues to 23. The newcomers are Bulgarian, Croatian, Czech, Danish, Finnish, Hindi, Norwegian, Polish, Romanian and Swedish, as trumpeted here. All fine and dandy, but just how does Google Translate perform …
I thought I would give it a go, so tried translating a few phrases and pinged them to my Czech girlfriend. Apparently they were near-total gibberish and certainly bore no similarity to what I wanted to say.
It couldnt even manage to translate 'So you are not impressed with my Czech?'
which came out as 'so you are not my impression Czech?'
El Google habla muy bien el Espanol! El Lester Tambien!
Seriously, I have been using Babelfish & Google Translate for over a year now to do prelim translations into French and Spanish for web-site content. As I speak and read these to languages pretty well, I am able to correct the results quite easily. Not perfect but enough to be comprehensible and generate business.
Another trick to getting better results when machine translating from english to these languages, is to write the english version as it would be written if literally translated into the other languages. e.g. don't use apostrophes etc.
For example, using googletranslate from english to spanish, a coloquial english phase:
>> Lester's Siesta translates as Lester's Siesta
>> The Siesta of Lester translates more correctly in Spanish as "La siesta de Lester"
which is fine and dandy if, like you, you can speak these languages, and therefore can phrase the input language to get the desired result. Smartarse.
Unfortunately, the less linguistically talented majority of users (yes, I'd be one of those, smattering of school French / Spanish from MANY years ago) will just have to put up with the gibberish.
I'd like to see them translate "Late Saturday Night Drunken Glaswegian" into English.
Now THERE'S a challange.
I remember reading about Googles super-duper translator-ga-tron ages ago:
And it wowed folks with its ability to translate even languages like Arabic into something very readable. If Google Translate was the fruit of that project, when did the apples start tasting of potatoes?
As a professional (Czech->English) translator, I should point out that no machine translation can ever be as good as one done by a human being. Well I would say that, wouldn't I? But really, machines just don't get the nuance, much of the idiom, or most of the slang - and they really have a problem with technical jargon that isn't pirated from English in the first place. Throw flexible word order (of the sort that Czech has) into the mix, and the fun really begins!
But this doesn't actually matter.
Google isn't (I hope) trying to put every translator on the planet out of a job. They're just providing a tool to increase the usefulness of Internet pages. The idea is to get the gist of what the page is about, not to provide something suitable for printing. Darned useful.
This is a world away from those who market translation software that IS designed to put translators like me out of a job, and which despite the marketing fortunately suffers from all the same problems.
I read this and thought: "but Spanish isn't one of the new features, that's not a proper test!". So here's the same piece of text translated from Spanish to English via Bulgarian, Croatian, Czech, Danish, Finnish, Hindi, Norwegian, Polish, Romanian and Swedish, according to Google:
"Guardia Civil de Huelva 29 people were arrested yesterday and handle 4,000 kg per hašiš most measures against drug trafficking, which was formed in the province. There is considerable work has been divided into two separate and cooperation with the customs authority. 16 persons were arrested and the rest of Russia, Spain, Poland, Romania, Ukraine and Morocco yesterday in a national movement."
Yes, actually I *do* have better things to do.......
Google translator do not "guess" a translation for a certain word. It tries to match a whole phrase against another whole phrase, based on known translated documents, like documents from the United Nations and from the European Union. The quality of the translation relates very closely to the quality of the original texts, and whether the phrases exist on those documents. That quality will improve as more of such documents exist.
Translating to Welsh would be clever - although I could see some difficulties with the mutations.
Dog is 'Ci' in Welsh, but depending on whose dog it is, 'his', 'hers', and 'my' changes the word changes to 'gi', 'chi' and 'nghi' respectivley. House would change from 'Tŷ' to 'Dŷ', 'Thŷ' and 'Nhŷ'.
Also Wales is Cymru in Welsh, 'I live in Wales' is 'Dw i'n byw yng Nghymru' but 'I come from Wales' is 'Dw i'n dod o Gymru'.
There are other rules about mutations:
After ei when it means "her"
After a ("and")
After â ("with")
Masculine nouns after the number three (tri)
After the number six (chwech, written before the noun as chwe)
This makes it difficult to look up words in a Welsh dictonary because you may try to look up the word Dŷ but you won't find it under 'D' but under 'T' because it is mutated from the word Tŷ.
A similar thing happens in placenames. 'Swansea' is 'Abertawe' in Welsh, which means 'mouth of the Tawe (the river which runs through Swansea)'. A town just up the road is 'Pontardawe' which means 'bridge over the Tawe' but the 'T' of Tawe has mutated in to a 'D'
Surely the best way to test the translation is to translate from English in Foreign and then back again.
The only time I tried Google translation was for "Fish and chips" which became ( in Italian) "Pesci e Circuiti integrati , but today I found that the italian for "Fish and chips" is apparently "Fish and chips" so perhaps those foreigners have seen the light.
...back in the old Soviet days, the Russians wanted something that translated English/American into Russian. After a lot of effort, they finally got something and the first try at translating - out of sight, out of mind - came back as - invisible and insane !!
Back to the drawing boards, lads, unless you want a free holiday in nice Uncle Joe's Siberian resorts !!
It did "My name is Aditya Krishnan. I live in India" in Hindi pretty ok. Butchered the gender though (in Hindi verbs change according to the gender and number of the object). With the number of my compatriots working in Google, though, it should improve double-quick. Pretty good at French and German though. and when I tried to translate "genetic mutation" into Russian, it gave me something that turned out to be "Marathon, whence".
Of the languages that can be google-lated (or transgoogled), only five are non-Indo-European: Arabic, Chinese, Finnish, Japanese, & Korean. One Semitic, one Sino-Tibetan, one Uralic, and two that may, or may not, be related to the Altaic languages and thence to the Uralics.
They really ought to add Georgian, one of the widely spoken Bantu tongues, and for good measure an Australian Aboriginal or Amerindian tongue to see how well their approach works when put to the acid test. Oh, yes, Basque would be a good trial too.
What's not obvious unless you look into it is that IE languages all have a strong family resemblance that goes a lot deeper than just vocabulary and some (nearly trivial) grammatical details. I've lately been entranced by Georgian and its gorgeous alphabet, and found some excellent web sites devoted to its grammar; the Wikipedia is quite good in fact
My head was spinning thanks to the concepts of "shreeves", and a highly agglutinative approach to verb forms, among other features of Georgian. When I woke up, it was with the insight that all IE languages are, in fact, much the same in their worldview, but other languages use totally different strategies to get the point across.
Regrettably, Georgian is not a very important language, but if any Google engineer is reading this, go ahead, give it a try. It really well test the basic concepts underly Google-lation.
I speak Swedish as a second language. I remember having to use an on line Swedish to English translator when starting to learn Swedish. The results were pants.
This seems to be a big improvement. I've used articles from several on line Swedish newspapers and the results in English are pretty promising.
Well done Google.
I have found Google's translator ok at German but useless at Italian and Spanish.
The obvious problem is that (the gender of) the subject pronoun of a verb is omitted in these two languages (whilst that of inanimate nouns and adjectives is stated repeatedly, eg "lA miA cartA bancariA e` arrivatA"). Less clealy explicable is that Google doesn't understand the use of impersonal pronouns, eg "se puede hablar Espanol" ("you can speak Spanish", but literally, "it can speak itself Spanish").
My Italian teacher once set us a passage from Italo Calvino, whose vocabulary is enormous, even though he came from Argentina. In order to avoid spending several hours turning over the pages of the dictionary, and knowing that Google couldn't actually render it into English, I looked up the passage on the web and fed it to the translator. I still had to look up a couple of hundred words!
On the other hand, just try reading something in Japanese without it!
On one occasion I wanted to save money on translation and asked a professional translation www.wolfestone.co.uk for a proofreading quote and sent them a document translated by Google. The agency noticed straight away that it was a poor machine translation and wasn’t even able to proofread it. They had to do the translation from scratch.
What Google have done is pretty impressive but at the moment it's a literal translation. It doesn’t translate idioms or technical phrases correctly. Also if you need a certified translation of your marriage certificate or business documents you need to go to a professional translation company.
And the English wonder why the Scots want to break off on their own, and the Welsh Nats want to do the same....
The french hate the English for some inexplicable reason (57% of the French population were killed in the 100 years war), the Irish hate the English (various small things like genocide by Oliver Cromwell, pitchcapping, the potato famine, black and tans dragging young men to their death behind police vans, Irish regiments being sent first to be annihilated by german machine guns in 1914-1918). Indians feel pretty much the same for the same reasons, and arabs the same (small matter of the crusades, plus continued interference upon the discovery of oil).
And the icing on the cake is the English are so ignorant that they know nothing of the above, and talk about how they brought civilization to the globe...
It seems also that the Englishman is set to be a thing of the past, with more people in England seemingly ethnic of some description.
Not all Englishmen are like this, but their collective "national identity" is far from something to be proud of.
"they are so ignorant that they know nothing of the above, and talk about how they brought civilization to the globe..."
trouble is, anyone acting like this will make you angry...
The french are more vocal about it, that is all...
If you TRY to say a few words, no matter how badly, they will be
very pleased that you TRIED... and usually will reply in perfect english!!! :)
>>if you TRY to say a few words, no matter how badly, they will be very pleased that you TRIED... and usually will reply in perfect english!!! :)
Um, not in my experience. Whenever I've tried (not completely ineptly) to speak French to the French, they have looked at me with such utter disdain that it's like a whole new level of disdain like you've never seen before. But then I know Paris is 'different'. (If anyone posts that icon after this I will punch them to the ground.)
My friend was going to learn how to perfectly say, "I shall not sully your beautiful language with my clumsy tongue, please may we converse in English?" Well, it's got to be better than pointing at pictures in a book and doing Looks and muttering "mercy bo coo".
Biting the hand that feeds IT © 1998–2020