In our last tip we mentioned the three major problems of automatic translation, particularly the new so-called statistical translation, which you get for free in countless Internet sites.
Statistical translation is based on giant databases with millions of pre-translated sentences. When a new text is entered by the User, the computer compares it with similar texts already existing as pre-translated in its database. However, the new text sometimes has to be identical, using exactly the same words and the same grammatical construction, to elicit a result. A slight variation could compromise the whole process. The three problems that we referred to in our last tip were:
How do you know which sentences are already included as pre-translated in a database?
If you get a bad translation, how do you fix it?
How do you know if you are getting a bad translation in the first place, if you are not fluent in the other language?
Here is an (extremely) elementary example:
- Buenos días, profesor = Good morning, teacher. (Perfect translation!)
Now, let’s change the grammatical construction slightly:
- Estos son buenos días, profesor = These are good morning, teacher. (X)
Of course, if you only speak Spanish you have no way of knowing if the English rendition is correct. Then, if you take the English translation and copy-paste it back into the statistical translator, to “test” the results, you will surprisingly get a correct retranslation, which will deceive you into thinking that your English translation was “perfect”.
Let’s try it:
- These are good morning, teacher = Estos son buenos días, maestro
What happened here was that, by a weird coincidence, the back translation was wrong also, and the second mistake compensated the first one. Click Here to download a free trial of our Text Translator.
Don't forget to stop by our newly-added Translation & Grammar Forum
to check it out and dispel all your doubts
on translation and grammar issues!