Machine Translation

While checking my stats, I noticed that the visits to my site had gone up sharply over the weekend and that several visitors had googled my full name. Now that could’ve been anyone, an old classmate, distant relative, estranged friend, potential customer. Nonetheless, I wondered whether it wasn’t one of my students. I don’t normally use my first name at school, but it’s not exactly a secret either, so a student might have found out.

Now I am deliberately vague about my school work. I never name the school, let alone individual students. I sometimes recount stuff that happened at school, but I never mention anything that’s potentially embarrassing to the participants. Plus, my blog is in English, so the students won’t be able to understand more than a word here or there. I teach English, after all, so I know the extent of their language skills.

Anyway, I suddenly wondered whether the kids knew about Google Translate or Babelfish or similar automatic translation services. They’re not exactly difficult to find after all. So I pasted a blog entry about school stuff into Babelfish and Google Translate. The output from Babelfish was completely unreadable, though very funny. Google Translate was at least readable, though still pretty bad. Interestingly, Google translates “arseholes” as “Esel” (literally “donkeys”, figuratively “idiots”), while Babelfish uses the more accurate “Ärsche”. And no, the result wasn’t overly revealing either, not that I would have expected it to be. Most of the time, I only blog about books and films and writing anyway.

You occasionally hear claims, usually from the overly technophile pages of Wired, Scientific American or magazines like that, that computers will be able to fully replace human translators within five years or so. For some reason, whatever total and radical change these mags are promoting is always exactly five years away, just as global oil reserves will always run out in exactly twenty years and have been since the 1960s, which is a wholly different rant.

But while Google Translate and Babelfish are certainly an improvement compared to early attempts at machine translation – and machine translation was among the earliest applications attempted to develop for computers, at a time when computers were still the size of warehouses and the PC, let alone the internet or Google, were not even a mote in anybody’s eye. Nonetheless, one thing hasn’t changed since those very early attempts in the 1960s and that is that machine translation just isn’t very useful outside very strictly limited contexts (one of the few successful attempts was a Canadian program to translate weather forecasts from English into French and vice versa) or for a very rough overview, i.e. what Google Translate does. If you want translations with correct grammar, respecting linguistic nuances and free of unintentional howlers, you still need a human and will continue to need one for the foreseeable future.