A new target language for machine translation
« previous post | next post »
Weasel-speak, as featured in today's Tank McNamara:
There's clearly money in it — and quite a bit of training material out there.
October 18, 2009 @ 9:51 am · Filed by Mark Liberman under Computational linguistics, Linguistics in the comics
« previous post | next post »
Weasel-speak, as featured in today's Tank McNamara:
There's clearly money in it — and quite a bit of training material out there.
October 18, 2009 @ 9:51 am · Filed by Mark Liberman under Computational linguistics, Linguistics in the comics
Powered By WordPress
Peter Taylor said,
October 18, 2009 @ 10:08 am
Would that be the first case of machine translation in which the goal is not to preserve the semantics?
DYSPEPSIA GENERATION » Blog Archive » A new target language for machine translation said,
October 18, 2009 @ 11:30 am
[…] Read it. […]
D. said,
October 18, 2009 @ 1:59 pm
Oh, that's delicious.
Tank McNamara is very occasionally quite funny, although less than it used to be.
Morten Jonsson said,
October 18, 2009 @ 2:26 pm
The cartoon is cute, but it's well understood that players' lawyers are the ones who write those statements. A cartoon of an athlete speaking to the press with his lawyer interpreting for him would be truer and funnier.
David said,
October 18, 2009 @ 2:52 pm
We really need to use Weasel-speak as the source language. Unfortunately every domain has its own method of Weasel-speak.
Theo Vosse said,
October 18, 2009 @ 3:35 pm
I have the vague idea most of these statements translate quite literally, actually. E.g. "This is very unfortunate personal situation" can be directly translated to "Dit is een zeer ongelukkige persoonlijke situatie" (reader game: guess the language). The problem would seem to be to *preserve* the ambiguity, rather than resolve it as is the usual goal in (can I already say: classical?) MT. E.g., "my children that I know about" sounds to my non-native ears rather vague, avoiding but not concretely admitting having illegitimate offspring, while the direct translation would leave no doubt at all.
N said,
October 18, 2009 @ 4:36 pm
It seems that there would be little correspondence between the two semantics of Regular versus Weasel speech. Maybe one would want to (if this were serious) do a finite-state system where if something occurs in a sentence, like "it's not my _fault_", it would trigger an entire sentence about culpability:
[s]…fault…[/s] –> This is a very unfortunate situation that has resulted from my actions, and I sincerely apologize for my behavior.
Might be fun. See the correspondence between news reports concerning an event and the press releases by those involved.
Richard said,
October 18, 2009 @ 5:53 pm
This is not a new topic for comedy; in this extract from the ever-funny 1980s British sit-com set in 10 Downing Street, the Prime Minister, James Hacker, and his private secretary, Bernard Woolley, are composing a press release in response to claims that that Hacker attempted to suppress the publication of his predecessor's memoirs:
(Yes Prime Minister, 'Official Secrets', 7'30" in)
I wonder what antedatings might be found for this phenomenon.
Levi Montgomery said,
October 19, 2009 @ 10:04 am
Not quite able to make out the text in the panels, I do a rollover to see if it's a link. "Click to embiggen," it says. Nice!
speedwell said,
October 19, 2009 @ 11:20 am
I'm going to guess Afrikaans?
roscivs said,
October 24, 2009 @ 10:46 pm
Nope, the Z makes it Dutch (or maybe Flemish?).