It's somehow appropriate that in the same week that Michael Scott leaves The Office that two computer scientists at the University of Washington announce their particular breakthrough in natural language programing — a computer program that knows when to say "That's what she said" to a double entendre.
It sounds like a waste of time, but this type of language play is actually a complicated computer science program. As the researchers note, this is actually a pretty tough problem, given that the natural ambiguity and cultural context of the language both have to be taken into consideration.
The system the researchers used to program their "TWSS" program is called Double Entendre via Noun Transfer, or DEviaNT for short. It's basic problem to solve is to be given a sentence, then determine whether it's funny if you add "That's what she said" to the end of it. To "train" the computer how to identify the right sentences, they used twssstories.com, a repository of user-submitted "TWSS" jokes, and non-entendre text from sites such as Wikiquote.
In the final results, the DEviaNT system knew when to say "That's what she said" about 72% of a time – an excellent achievement for this type of natural language program. And frankly, I know people who don't know when to say "That's what she said" correctly at that high a percentage, so I'm doubly impressed. However, the researchers are convinced that given a better sample set, DEviaNT would totally nail it.
That's what she said.
Thanks to Alex Knapp / Blogs Forbes
No comments:
Post a Comment