To a certain extent, it’s quite amazing that a piece of software can take a piece of text and magically translate it into any languages. Website owners often wonder how accurate those free tools are and if they should use it on their websites. Unfortunately, the result is often dismal. Being native in French, and fairly at ease with Spanish, I always notice a website that has been translated using a free machine translation tool, and it is not pretty.
Now, being that I represent a human translation agency, I’m probably not exactly objective when it comes to machine translation. So, instead of taking my word for it, let’s take a look at a simple example.
Take the below short four sentences paragraph, which I wrote in French, and then translated myself (literally) in English:
Bob travaille pour une petite entreprise au centre-ville. Le travail de Bob consiste dans la réparation d’ordinateurs et l’installation de nouveaux logiciels. Durant ses temps libres, il aime lire, regarder des films et jouer au poker avec ses amis. Bob a 30 ans et vit dans une grande maison avec sa femme et ses enfants.
Bob works for a small company downtown. Bob’s job consists of repairing computers and installing new software. In his spare time, he likes to read, watch movies and play poker with his friends. Bob is 30 years old and lives in a big house with his wife and kids.
Now, I used one of the popular free translation software to translate the same French text. I got the following result:
Bobsled works for a small business to the center city. The work of Bobsled consists in the repair of computers and the installation again software. During its free time, it likes to read, look at films and to play poker with its friends. Bobsled is 30 years old and lives in a big house with his woman and its children.
Four sentences, and already we can spot at least 10 significant mistakes. And don’t ask me why Bob gets translated in Bobsled!
So, let’s take it further. Since many use free machine translation to offer their pages in other languages, let’s try to get a better idea of what get shown to their foreign visitors.
I took the human translated English text and used the machine translation to translate it in French and then retranslate it in English. Here is the result:
Bobsled works for a small company in city. The bobsled work consists in to repair computers and installing again software. In its leisures, it likes to read, look at poker of films and piece with its friends. Bobsled is 30 years old and lives in a big house with his woman and its childlike ones.
Obviously, here, there’s the double effect of the machine translation, but, truth is, machine translation is even worst when translating from English into another language than the other way around. Publishing copy like the one above is like slapping your readers in the face!
I just tried translating the French paragraph using Google Translate and the results are amazing:
“Bob works for a small business in downtown. Bob’s work consists of repairing computers and installing new software. During his spare time he likes to read, watch movies and playing poker with his friends. Bob was 30 years old and lives in a large house with his wife and children.”
There are a few minor mistakes “playing poker” instead of “play poker”, “a small business in downtown” instead of “a small business downtown”. and “was 30 years old” instead of “is 30 years old” which can be corrected in a couple seconds. I don’t speak French but could quickly correct these errors. If you compare to the human translation, even the human translation has a mistake – “consists in” instead of “consists of” and I’m sure it took much longer.
worldbeing – Thanks for dropping by, (and thanks for spotting that mistake (mine) in the Human English version)! I’ve never tried Google Translate, but from your result, I must admit that it’s more accurate than the other machine translation tool I used in my example.
However, in your example, you used the tool to translate into a language you know. There’s a problem when webmasters use these tools to offer their website in other languages (that they don’t speak). Had you translated this paragraph into Russian for example, assuming you don’t speak Russian, you never would have been able to spot the few minor mistakes nor correct them in a couple of seconds. Now, imagine a complete website, filled with minor mistakes. In addition, the text in this example is as simple as it gets, so a few mistakes is probably a best case scenario.
–Charles
I have tried google’s English-Russian translation and vice versa and it is not so bad. As was mentioned above, even humans make minor mistakes and the idea of free MT software is not to provide a perfect translation but to give you an overall idea of what the text is about in a pleasing output. That some people try to save money by translating everything with machines is a question for another discussion. Human translation is the best solution but it is slow and expensive. You would agree that when you need a short paragraph or a few pages from a report on the Internet you can’t always go to a translator and pay a lot of money and wait a day or two. The same is with owners of websites. When a site is starting owners not always can afford effective HT and if they want the site translated into more than two languages and have to pay for every changed sentence in the text….the amount of money is huge.
This does not mean that I completely disregard HT, I am just saying thet machines are getting better in translation but one must be careful what they use them for. It can’t be expected that a machine can translate Homer or Dostoevsky but they do well on website translation and text with normal sentence structure (not a reduced one or one with missing information or ambiguities). There are some new products which work quite well. There is google which is free, MozTrans by Interlecta which is also free bu the same company has some paid products which are really good: http://home.interlecta.com/products
Continuing right along in the vein of the athor right above me: MT is not supposed to replace HT. It is right now a tool that enables fast HT via post-editing. Since the volume of text to be translated seems to be only increasing, MT can be very helpful in increasing productivity. That’s what it is intended to do.