trans.jpg

Translation can be a very entertaining business. I was once asked to “clean up” a text that had been machine translated (that means computer translated for those of you not in the biz.) from French into English. It was an RFP for upgrades to a coal-fired power plant in Morocco. (How entertaining could that possibly be? You’d be surprised.)

The company I was working for only had a hard copy of the original document so, before they could run it through the translation software, they had to run it through OCR (optical character recognition) software. Basically, you scan a doc and OCR software recognizes the shapes of the letters, punctuation, etc. and saves the file as editable text rather than an image file.

That’s always fun. In the case of my project, the OCR software often read the letters “ril” as “n” so, for example, “risques et périls” (a stock liability-related term in French) came out as “risques et pénis.”

Then, when they ran the OCR gobbledygook through translation software, this is the kind of thing they got (which I had to clean up):

  • it will not be tolerated that provisional constructions drink some such as cloakrooms.
  • The CONTRACTOR ensures in particular his expenses, risks and penis, transport and all the operations being attached to it (handling, storing, etc…).
  • Le-CONTRACTANT will have to make its deal of the telephone food of its building site.
  • The CONTRACTOR is held to make the desired provisions so that a doctor can quickly arrive on the spot and that the broken workmen can be evacuated quickly. He will possibly have an ambulance.

Hunnnndreds of pages of this stuff. And they wanted me to spend no more than 20 minutes on each page. If only people had a clue. Sigh.

I was inspired to write this because of a post on TechBee’s blog about a book on funny translations. The book has a section on priceless bad translations, like this one, which was originally in French:

Competitors will defile themselves on the promenade at 11am and each car will have two drivers who will relieve themselves at each other’s convenience.

Those French are such libertines! Shocking!

Clearly a translation job in the wrong hands can be a dangerous thing. I mean, an entire big, bad religion is based on a less-than-immaculate translation. Too bad humanity can’t prosecute that translator for the damage he did…

I don’t think human translators are an endangered species quite yet. But I bet Vincent wishes some computer could take over for him right about now. As I write this, he’s busy translating 75 thousand words of this into French:

The polymer mechanical property data is not specified. Residual stress and shrinkage and warpage analyses require the polymer mechanical property data to be specified. Check the material data, define all the relevant mechanical property data, then re-run the analysis. For isotropic materials, define the elastic modulus (E) and Poissons ratio (v). For transversely isotropic materials, define E1, E2, v12, v23, and G. For other supported viscoelastic models, define constants.

You gotta be a special kind of crazy to take a job like that.

If you’re a language geek and you want to have some fun with machine translation, go to Babelfish. Paste some text into the window and translate it. The more unusual the original text, the better. The result might be mildly funny. But it gets really good when you take the resulting translation and translate it back into its original language. Do you recognize this song?

In the city where I was born, lived a man who sailed with the sea, and it told us his life, in the ground of the submarines, Ainsi we sailed until with the sun, until so that we found a sea of green, and we lived under the waves, in our yellow submarine, We all live in a submarine yellow, underwater yellow, underwater yellow, us that all live in a submarine yellow, underwater yellow, underwater yellow, And our friends are all on board, much more them next door of phase, and the band starts to play.

I actually kind of like the abstract, poetically incorrect result…

I worked briefly on dictionary development for Systran (the company that created the MT software used by Babel Fish), on the Alta Vista/Babel Fish project before it was launched. So if you play with it, some of the words that come up might actually have been entered into the dictionary by my very fingers!

More on how Systran’s MT software works for you über geeks. And check out Roger’s blog on all things linguistic!