Is Automatic Language Translation possible?

[minttwist]

I'm sure of one thing; language translation for the web is soon going to become a huge sector of the web development market.

Our web agency is being asked more and more to create multi-lingual websites. We are outsourcing the work (to agencies) and in some cases businesses are spending 10's of thousands on translation (as much as the website creation itself).

We have been looking at Google translate as an option to put to customers on a budget. The software is plugged into chrome and can be used to automate translation.

Unsurprisingly, the big translations agencies argue against this:

http://www.todaytranslations.com/blog/784/is-automatic-language-translation-possible

and technical agencies are for it, pushing solutions like this:

http://labnol.blogspot.com/2005/11/add-language-translation-to-website.html

I'm not sure where it will go - but my data is showing me that one way or another there is going to be huge sums of money to be earned for businesses who can provide localisation and translation at value prices.

 

[ivebeenstiffed]

Aoccdrnig to rscheearch at an Elingsh uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are, olny taht the frist and lsat ltteres are at the rghit pcleas. The rset can be a toatl mses and you can sitll raed it wouthit a porbelm. Tihs is bcuseae we do not raed ervey lteter by ilstef, but the wrod as a wlohe.

 

[minttwist]

Meaning can be decoded, but without care it can go wrong. Especially with language translation.

Below are some real life examples of when translations go wrong, including the sign put up by the big Chinese Hotel that were trying to be helpful - but ended up being rather intimidating: "Take luggage of foreigner - no charge"

http://todaytranslations.wordpress.com/2010/06/02/english-translation-gone-wrong-part-2/

 

[maxh]

Tell people to use google chrome.

It automatically translates most text on a website

Try it for yourself foo'

Chinese Website

 

[KidsBeeHappy]

I think the general status is that words can be translated automatically, sentiment can't.

So, if its things like forms, or instructions, then yes I would probably at least consider an automated translator. But for something like T&C, any legal document, any tender or proposal etc, then that touchy feely thing I think still needs a human to interpret the sentiment.

 

[maxh]

I think the general status is that words can be translated automatically, sentiment can't.

So, if its things like forms, or instructions, then yes I would probably at least consider an automated translator. But for something like T&C, any legal document, any tender or proposal etc, then that touchy feely thing I think still needs a human to interpret the sentiment.

Yes, and this^

 

[Digger]

My wife ( Russian ) did some translations for a company - it is in fact a VERY time consuming business to do it correctly - especially as Technical Terms rarely translate easily - and often there is no equivalent.

I think it will be a time before a software programme can do it

 

[SusanP]

Machines aren't sophisticated enough to recognise subtleties in language. For instance, 'by the sea' and 'by sea' mean different things, as do 'what is time?' and 'what is the time?', 'a few' and 'few'. Russian has no definite or indefinite articles (the, a/an) so an English to Russian automatic translation tool could not cope with those sentences.

 

[the_agx]

Lo vengo diciendo desde hace rato.

The above is Spanish. It translates to: "I've been saying it for a while". But if you put that in a Spanish-To-English translator:

- you are likely to get each WORD translated in turn, rather than the whole phrase analysed.
- the semantics, the meaning, subtleties, etc, not translated well.
- nonsensical robotic machine-like translation.

Something like:

"Him I come saying since it makes while". (if the translator were crude and amateurish, I suppose)

"Hace" from the verb "hacer" (to make/to do) means "it or he or she makes/does". But it has so many idiomatic meanings. With expressions of time, it roughly translates to "ago/for a [time duration]". How do you know when to translate "hace" as:

Lo hace bien = It s/he does well = S/he does it well.

or as:

Hace mucho tiempo = It makes a lot of time = long time ago.

If you didn't understand the above example, don't expect a computer to do so either!

 

[minttwist]

I agree that Google Chrome is impressive as an automatic translator. However, there is no doubt that sentiment and some meaning is lost...

Moreover, if we are talking about trying to communicate very clear and specific messages to a specific set of (non-English speakers) - something which most business websites that are pitching their products/services to foreign markets are trying to do - then asking the potential customer to "Please always use Google Chrome" is not a great start!

 

[dagr]

I remember in the 1970s being told by serious programmers that automatic translation was just 5-10 years away. Ho-hum.

It's easy to check how bad automatic translators are by simply entering a few simple sentences and then going back and forth between the original and target languages. It will almost always end up as grammatically correct gibberish nonsense compared to the original text. Beware.

 

[minttwist]

I remember in the 1970s being told by serious programmers that automatic translation was just 5-10 years away. Ho-hum.

It's easy to check how bad automatic translators are by simply entering a few simple sentences and then going back and forth between the original and target languages. It will almost always end up as grammatically correct gibberish nonsense compared to the original text. Beware.

Agreed.

A client of ours paid £150 to have a single landing page professionally translated into Spanish. In the last 4-months they have two projects from Spanish businesses worth a combined £25k. I do not believe Google Chrome could have the same effect. Previously they had 'website hits' from Spanish businesses but had never won any business. They engage the potential customer in their own language on the website, however, all follow communication is in English.

http://www.conflictinternational.co.uk/Detective-Privado.html

They have subsequently been quoted only £750 for the whole site! If UK SME's are going to export the country out of debt this looks like a cost-effective way to start.

At those sort of prices and with that effect, if a translation job is worth doing, the professional agency still provides massive benefits over an automated solution.

The translations agency was http://www.todaytranslations.com

 

[efece01]

Google translator is quite nice to get each word translated.
By setting up a sentence, it gives you an idea of what s the text about.
An idea, not a translation.
It is not a translation service but a way to understand other language websites.

 

[maxh]

Google translator is quite nice to get each word translated.
By setting up a sentence, it gives you an idea of what s the text about.
An idea, not a translation.
It is not a translation service but a way to understand other language websites.

Google translator is about to get much more powerful with the addition of conversation translator!

 

[Raw Rob]

Lo vengo diciendo desde hace rato.

The above is Spanish. It translates to: "I've been saying it for a while". But if you put that in a Spanish-To-English translator:

- you are likely to get each WORD translated in turn, rather than the whole phrase analysed.
- the semantics, the meaning, subtleties, etc, not translated well.
- nonsensical robotic machine-like translation.

Something like:

"Him I come saying since it makes while". (if the translator were crude and amateurish, I suppose)

Automatic translators have come on a long way in the last few years - if you try that example in Google Translator, the result is:

I've been saying for some time.

Rob

 

[Astaroth]

Google translator is about to get much more powerful with the addition of conversation translator!

Is this even new? New from Google obviously but there are plenty of long standing technologies that do both text translation and text to speech. There is also plenty of dictophone type systems that do speech to text so it wouldnt be hard to glue these technologies together and create what google is offering. I'm sure I;ve seen it before already.

As others have said, the issue is that technology is very good at translating the words but not the sentiment. Even if you put simple sentances through an online tool then translate it back again you can start to see the issues it generates as the back translation is rarely the same.... normally close enough to get the sentiment but not the same. If this is your business getting "close enough" really shouldn't be exceptable if your targeting people who speak that language.

Add more complexities in like idioms and you can totally lose someone because not only do they not translate on a word for word basis but even different countries that speak the same language tend to have different idioms for the same concept.

Finally you then have to consider culture, what appeals to one population isnt going to appeal to another.... if your website is pushing your products hard on the basis they're made locally in the UK, low carbon footprint etc etc, you cannot just translate that into portuguese and think they will sell well in Brazil. You then need a copywriter as much as a translator.

Technology can go a long way in helping to translate but is a clear second best to a human translation.

 

[maxh]

Is this even new? New from Google obviously but there are plenty of long standing technologies that do both text translation and text to speech. There is also plenty of dictophone type systems that do speech to text so it wouldnt be hard to glue these technologies together and create what google is offering. I'm sure I;ve seen it before already.

As others have said, the issue is that technology is very good at translating the words but not the sentiment. Even if you put simple sentances through an online tool then translate it back again you can start to see the issues it generates as the back translation is rarely the same.... normally close enough to get the sentiment but not the same. If this is your business getting "close enough" really shouldn't be exceptable if your targeting people who speak that language.

Add more complexities in like idioms and you can totally lose someone because not only do they not translate on a word for word basis but even different countries that speak the same language tend to have different idioms for the same concept.

Finally you then have to consider culture, what appeals to one population isnt going to appeal to another.... if your website is pushing your products hard on the basis they're made locally in the UK, low carbon footprint etc etc, you cannot just translate that into portuguese and think they will sell well in Brazil. You then need a copywriter as much as a translator.

Technology can go a long way in helping to translate but is a clear second best to a human translation.

You're correct, however the manor that google approach translation is dynamic. i.e. the more you use it, the more it learns. It uses statistical translation now. By simply increasing the number of people that use it, refine it and teach it it will become MUCH better at translation.

I agree, it will be a long long time until a machine can match up to a human. But I would say that this is the best I've seen yet.

Just using Google's voice-search, I'm damn impressed!

 

[directmarketingadvice]

Lo vengo diciendo desde hace rato.

The above is Spanish. It translates to: "I've been saying it for a while". But if you put that in a Spanish-To-English translator:

- you are likely to get each WORD translated in turn, rather than the whole phrase analysed.
- the semantics, the meaning, subtleties, etc, not translated well.
- nonsensical robotic machine-like translation.

Google translate has it as "I've been saying for some time".

Seems ok to me.

I tried translating "to buy a pig in a poke" into French and it gave "acheter un chat dans un sac", which is correct (and not at all literal).

Steve

 

[Astaroth]

You're correct, however the manor that google approach translation is dynamic. i.e. the more you use it, the more it learns. It uses statistical translation now. By simply increasing the number of people that use it, refine it and teach it it will become MUCH better at translation.

I agree, it will be a long long time until a machine can match up to a human. But I would say that this is the best I've seen yet.

They are miles better than they used to be, though still not seeing many (if any) giving regional variation - try translating "do you like my bum bag?" -v- "do you like my fanny pack?" (the same question but with american or english wording) and its translated into a different spanish phrase. A concern of a statistical methodology is that the most numerous will start to win and the less knowledgable will start sending Mexican wording to Spain.... my misses, a native spanish speaker, cannot read 50% of a menu in Mexico because so many of the words for food are regional.

 

[WorldToWriters]

As a translation agency, I just wanted to present you with some of the latest industry news and some of the self-evident conclusions:

- Automatic translation is called Machine Translation in the industry, or MT
- Agencies have started to offer a cheap "MT+edited" translation of documents alongside their full human translation offerings - buyer beware
- Studies* show that over 70% of consumers surveyed globally are more likely to buy a product with information in their own language. And over 70% of consumers spend most of their online time on sites in their own language
- The language services industry is growing rapidly and is expected to reach $25bn by 2013

The quality comparison between MT and human translation is still not worth drawing - there is next to no way you could sell a product in a foreign language with MT yet. It is improving, through masses of statistical analysis, but there are so many nuances that it cannot yet cover that it cannot be considered a free ticket to foreign sales.

So for now, people are stuck with outsourcing or employing human translators/writers to convey their message. Depending on how much they value foreign custom, this could go without saying. And the cost really isn't as expensive as you might think, especially given the potential returns.

Part of the problem in the misconception of how useful MT is, is that when you translate something into English using, say, Google Translate, the result isn't terrible. It is certainly comprehensible. However, you will get different results in different languages. Less common languages give poor MT results, in general. Another problem is that there is no way to check the quality if you don't speak the target language. You may accidentally offend everyone who visits, at best by giving them sub-par text to trawl through, at worst by using offensive words. This won't help your sales efforts much to say the least.

Machine translation is often so difficult to read that people could be forced to just give up. This is especially true in some languages where literacy is highly regarded, such as French (or English for that matter); where a single mistake will see many readers switch off immediately.

So, as may be predictable from a translation agency (but I hope I've explained why), we would like to encourage you to have your text translated by native-tongue, trained professional translators, with experience in your industry for those specialist terms. Your clients will appreciate it many times more than the automated alternative, which may end up having the opposite effect!

We have also tested the leading supplier of paid-for MT software's offering and found it to be unusable, and that at costs going into the hundreds of pounds.

Given all of that, we have placed ourselves in a position to help. We offer a free report on how to get started in exporting - including how translation services can help, of course - on our website.

*Source: Common Sense Advisory, language industry researchers

 

[3Atech]

We cannot reliable software translation only, try to learn more.