When the result is a lousy paragaph
John Dvorak, who I recall as being pompous but funny at the Microsoft Windows 1.0 release event (yes, I’m that old) — and rarely funny thereafter — has a rant on machine translation. The second page has some examples of different translations of the same passage. They do make me wonder why translations aren’t run through a more complete grammar-fixing process. Or, if that would have too many risky side effects, why the same translation program doesn’t offer several DIFFERENT translations, tuned to optimize against different kinds of error.
The same principle, of course, could be applied to gists/summaries in other text technologies, such as search. Sometimes, a decent UI would have room to allow user configurability and/or multiple bites at the output apple.
Comments
Leave a Reply