As noted in the article, brain centers associated with attention are deactivated during sleep, so acquisition of semantic knowledge seems unlikely, as you suggest.
The experiment's novel condition was to begin a procedural task before sleep then continue the task as subjects fell asleep and slept, monitoring them the whole time.
This does remind me of a Peanuts comic that my 6th grade teacher had up on the cork board. It had Charlie Brown putting his school book under his pillow, and he said he was going to study for his exam through osmosis.
