Good grief, I wasn’t hallucinating.

Last night, around 3 AM, I had some problems falling asleep, so I visited a site about a rather famous Finnish cult. (I’m not discussing the cult itself, but if you want background information in English, try the Wikipedia article).

Now, the site itself is built on Google Sites. Naturally, the site owners have helpfully put in a Google Translate widget. Except, um…

Where's English?

…it’s pretty awesome that the widget lets you translate between Finnish and some other languages like, uh, Maltese and Haitian Creole. It’s also good to note that Chinese is supported - there are a lot of Chinese speakers in this world, you know. But there’s one pretty big omission. Hmm, wonder what it is… This big language thing… Hmm… Oh yeah, English.

And the widget lets you translate Finnish to… uh, Finnish???

Here’s a paragraph it took:

Toinen tarkoitus on antaa tietoa julkisuuteen siitä, minkälaisesta kuviosta on kysymys. Lahkot, jotka vahingoittavat ihmisiä ovat aina erittäin valitettavia ja niiden toiminta pitäisi saada loppumaan. Tapani Koivuniemen Konserni on tehnyt matkan varrella monia ihmisraunioita, joista useat ovat joutuneet jopa psykiatriseen hoitoon. Monilla menee useita vuosia toipua hengellisestä hyväksikäytöstä, mitä ovat Konsernissa joutuneet kohtaamaan.

And here’s how it’s translated to Finnish:

Toinen tarkoitus on antaa on tietoa julkisuuteen siitä, minkälaisesta kuviosta on Kysymys . Lahkot, OTTAVAT vahingoittavat ihmisiä ovat Aina erittäin valitettavia JA niiden TOIMINTA pitäisi Saada loppumaan. Tapani Koivuniemen konserni on tehnyt matkan varrella Monia ihmisraunioita, joista useat ovat joutuneet jopa psykiatriseen hoitoon. Monilla menee useita vuosia toipua hengellisestä hyväksikäytöstä, Mita ovat Konsernissa joutuneet kohtaamaan.

Google Translate, is a service that RANDOMLY commas and capitalisation adds. Word order also weird sometimes. It also makes the text, to LOSE umlauts.

Okay, how does it translate it to, say, Russian? I don’t speak Russian much, but changes should be pretty noticeable, right?

Toinen tarkoitus на antaa tietoa julkisuuteen siitä, minkälaisesta kuviosta на kysymys . Lahkot, jotka vahingoittavat ihmisiä ovat Айна erittäin valitettavia JA niiden toiminta pitäisi Саада loppumaan. Тапани Koivuniemen Konserni на tehnyt ihmisraunioita varrella аммиака matkan, joista useat ovat joutuneet jopa psykiatriseen hoitoon. Monilla menee useita vuosia toipua hengellisestä hyväksikäytöstä, MITA ovat Konsernissa joutuneet kohtaamaan.

Тис ис нот веры муц лаик Рашшан, ис ит? “Aina” becomes “Айна”, “saada” becomes “Саада”. “On” becomes “на”, which only makes sense if the thing is somehow trying to interpret the page as English text.

OK, so the widget is useless. What if we feed the page to Google Translate directly? Surely the results are more awesome?

OK, let’s feed it.

The company had the money to make spiritual products, if Stephen Birch peninsula would not want it. Everyone can think, for what purpose he is actually collecting the money.

…”Stephen Birch peninsula”? …oh, right, Google Translate is translating the names too. What. I can understand if the sotware wants to tackle “Koivuniemi” and it spits out “birch peninsula”, because there’s no easy way a machine could tell the difference between a surname or a description of a geographical feature. (Perhaps it could do some guesswork based on the fact that it was capitalised and followed a given name.) But there’s absolutely no logic behind trying to translate the given names to culturally equivalent names in other languages. Why the hell did it try to translate “Tapani” as “Stephen”?

It gets better:

For example, the garden director, John Doe, has testified that since 1995 he has never questioned the Tapania.

Yes, the thing translated “Jussi Virtanen” as “John Doe”. Wikipedia explains a little bit about the context here, but basically, “Jussi Virtanen” is not an unambiguous placeholder name; Virtanen is the most common surname in Finland and Jussi is just a fairly common name. Mr. Jussi Virtanen here is just a person with a dreadfully boring name. (No offence intended to the thousands of people who have that name - I know my name isn’t the best possible either =) By contrast, “Matti Meikäläinen” is the name that most people would immediately recognise as an obvious placeholder name.

In summary, Google Translate is still fun.