|
|
|
|
|
by PhasmaFelis
996 days ago
|
|
This reminds me: some years ago, someone noticed that some translation engine, maybe Google's (vague enough for you?) advertised a whole lot of obscure languages but didn't work all that well for some of them: if you translated from English to whatever and back, you'd get mostly gibberish with a lot of oddly Biblical turns of phrase embedded in it. I got to thinking about that, and I came up with a theory: Machine translation of course requires a corpus of texts that exist in at least two languages, a Rosetta Stone. The Bible is the most translated book on Earth. For tiny or dying languages, the kind with some thousands of speakers at most, the Bible may be the only written work that's ever been translated into or out of that language. So the Bible is their entire training corpus, and it's not enough. |
|