Why Google Translate wouldn't be possible without Harry Potter

Anyone who’s ever used an online translation tool knows that it’s a rather imperfect art.  Take a given sentence, run it through the translator, and then translate it back to English.  The result will be at best a little garbled, but will usually be pretty much incomprehensible. Here is what happens, for example, when you run a sentence through Google Translate:

Spanish: Esto es lo que sucede, por ejemplo, cuando se ejecuta una sentencia a través de Traductor Google.

And back to English: This is the case, for example when the statement is executed through Google translator.

Clearly, professional translators don’t have much to worry about.  What is so interesting about Google Translate however, is that unlike other translation tools, it doesn’t actually deal with the meanings of words at all. Google translate doesn’t care about word meanings, syntax, or vocabulary. It turns out that there are only two things that Google Translate really cares about: Harry Potter, and the United Nations.

Rather than try and do any actual translating itself, Google Translate figures that someone else has probably already done the hard work for you.  Google uses its incredible computing power to trawl through the vast swathes of human translation work, and pairs your English sentence with a human-translated equivalent.

Google’s database for doing this is huge. Whenever you ask Google to translate a sentence, it draws on vast archives of translated text, including everything the UN and its agencies have ever done in writing in six official languages.

Essentially Google Translate is only as good as the human translation that has gone before it. It is built upon the millions and millions of human translators who first produced the texts that Google uses as its reference points.

This is why books like Harry Potter are so useful.  With translations in 67 languages, Harry Potter provides an excellent frame of reference for Google Translate to draw upon.  While there may be no recorded history of direct translation between Hebrew and Welsh, by running both translations through the hub of the original English text, Google can attempt a direct translation.

Because Google uses context rather than meaning, this can often result in some rather amusing translations.  As you can see, there’s still a fairly long way to go.

 

For more fascinating information about translation (and the source of this blog post) check out 'Is That A Fish In Your Ear: Translation and the Meaning of Everything' by David Bellos