By Iohannes Aurum, in 'English to Latin Translation', Oct 1, 2010.
I'm more shocked that someone thought it worth the effort to scan the whole thing electronically than that it happened to get published in a vanity press.
What I find interesting about Google Translate is that it translates Latin so badly, when it does a reasonable job of Welsh, Danish, Swedish. We tend to think that a more structured and predictable language should be easier for a machine to translate. But most modern languages are SO unstructured that it uses a completely different algorithm, comparing word patterns, and that doesn't suit Latin.
I was saying something to that effect recently, adding that Google Translate usually managed to make some sort of sense even out of inflected languages like German and Polish. Almost immediately afterwards I tried to use it to cut out the donkey work on a piece of Polish poetry, and it returned pretty much gibberish. I hadn't realised that it worked by comparing word patterns; that would explain why it appears to give up when confronted with a highly idiosyncratic piece of writing that uses atypical word order. By contrast, it can handle reports of politicians arriving at places and saying banalities in its metaphorical sleep, because they're all more or less the same thing.
At the risk of introducing politics onto a learned discussion, I can't say I'm surprised! It just seems odd that the way we parse Latin (or Greek) is to a large extent mechanical. Spot the ablative, etc. It must be far easier for a machine to work it out - clearly a blended approach of the two is needed. Fortunately, my co-forum colleagues obviate the need for machine translation!
I agree! Google translate for Latin is simply awful...
"Homo homini lupus est" translating to "no free lunch" was funny for a while, as well as some other oddly fitting things Google Translate did.
Then I ran out of ways to play with it and it lies abandoned in my bookmarks bar, waiting for me to find one of its horrific Russian translations amusing again.
That being said, I wonder if somebody who's a better linguist than myself could actually write a constructed language with the aim that Google Translate's translation algorithm will return perfect results, just so that we can have one thing it doesn't fail at.
But they're not equivalent statements. Apart from which, I've just tried it, and it returned 'man is a wolf'.
It may have done so in the past, which I took to be FloridPansy's implication. It actually updates itself, though not always to a better outcome.
That was the implication, yes - Russian GT translates a bit better than it did a month ago, to the point where "zhopa" (ass) doesn't come up any more upon searching for 'fistula'. Of course they're not really equivalent statements, but I do have the right to a sense of dumb humor. XD I am a clever teenage girl, but still a teenage girl, and I do apologize for anything I screw up on here.
Google Translate has an advantage with living languages in that there's a constantly updating corpus of texts from native speakers it can easily avail itself of. Of course there's nothing similar for Latin, but unfortunately it still seems to operate on the same principal and consequently picks up all sorts of bad, amateurish Latin as if it were the equivalent of the output of a Cicero or a Caesar. This is one of the primary reasons that it will inevitably remain a failure.
But Google also picks up all sorts of bad, amateurish English as if it were the equivalent of the output of a Twain or a Swift. It doesn't cause a problem, right?
I think the problem is that there's a much higher proportion of good (or at least decent) modern English to bad modern English, than there is of good modern Latin to bad modern Latin.
I doubt the problem could be ſolved by better ſample texts alone.
That's true. I have the feeling its programmers didn't really know Latin themselves.
Looks like Yardex has created another way of providing wrong Latin translations.
The boys love the girl pueri amore puellae
Quoth Imber Ranae: 'it still seems to operate on the same principal and consequently picks up all sorts of bad, amateurish Latin as if it were the equivalent of the output of a Cicero or a Caesar.'
A reporter for the Austin American-Statesman made up some bogus Latin for 'hairy eyeball'; Google now translates that phrase,
'morem pellis hispidus distentione nervorum' as hairy eyeball, solely on that 1 source.
Unfortunately Imber has eloped with some turtle and hasn't been seen for quite some time...
Yes, I rather miss him
As do I.
Our Latin forum is a community for discussion of all topics relating to Latin language, ancient and medieval world.
Separate names with a comma.