(no subject)

I'm participating in an evaluation effort for machine translation evaluation.

I'm very impressed by the state of the art. Of course, this material is operating in-domain, and I recognize that there is a long way to go, but the current systems do surprisingly well, surprisingly often.