Humans doing translation the machine-way

If this thread deserves a special tag, forgive me because I really had
no idea under what if at all it ought to be classified.

A little background to this post:
I was browsing through past mails I'd received from the list, and
chanced again upon the thread regarding machine translation (where I'd
posted a reference to an article in The Economist, if it makes the
memory clearer). It got me thinking-- if machines, armed only with a
distraction-free hyper-sped processor could manage to translate text
based only on statistical analysis and matching, how well would humans
perform in the same task?

What piqued my interest further was the coincidence that I was working
on translations for my conlang, Arithide (sample texts of which can be
found at http://wiki.frath.net/Modern_Arithide and which happens to be
pronounced like Aristide, the ex-Haitian president). How well might a
human interlinear the romanised Arithide text given the original
beside it?

Which brings me to:
If anyone is game, how about trying to interlinear some of the sample
texts that I've posted on the FrathWiki page? They range from short
(Pater Noster) to very long (an Economist leader on Russia), and
happily I haven't brought myself out of reluctance to provide
interlinears.

And to:
How about posting interlinear-less side-by-side translations of
certain texts in your own conlangs too, for the rest of the list to
try out?

Cheerio!
Eugene

Reply