I've been known to use this joke

Occasionally. First there was Watson, winning Jeopardy. Now there is a robot comedian. Cue the Terminator....From Katy Waldman, writing in Slate.

[Subject] Could Eat [Object] All Day. That’s What She Said.
If you have a shaky sense of comic timing (and you're a little immature), there's good news from the University of Washington. As the New Scientist's One Per Cent blog reported on Friday, researchers have developed a computer program that helps you identify the perfect opening for a "that's what she said" joke.

Computer scientists Chloé Kiddon and Yuriy Brun are interested in how humans recognize double entendres—and whether machines can learn to do the same. Spotting double entendres requires "both deep semantic and cultural understanding," they write (PDF). As Kiddon explained in an interview, a double entendre is really a type of metaphor that brings together two conceptual realms: one straight-laced and one raunchy. So "that's what she said" jokes aren't just crude, cheap ways to get a laugh-they're also fertile testing ground for whether computers can be trained to "think" metaphorically about language, the way humans do.

Kiddon and Brun define a TWSS as a sentence that is funny when followed by the phrase "That's what she said." Telltale TWSS markers include 1) the presence of nouns that are often euphemisms for more sexually suggestive nouns and 2) syntactical structures common to X-rated literature. The researchers give banana as one example of a seemingly respectable noun that could moonlight in porn writing. For racy syntax, they offer "[subject] stuck [object] in" and "[subject] could eat [object] all day."

To train their computer program, DEviaNT (Double Entendre via Noun Transfer), which assesses the TWSS potential of individual statements, Kiddon and Brun gathered 1.5 million sentences from erotic literature and 57,000 from more mainstream texts, such as Barry Goldwater's 1961 essay "A Foreign Policy for America" (which is just chock-full of euphemistic eroticism, we're sure). By analyzing big swathes of lexical content, DEviaNT began to learn which terms frequently appear together in risqué contexts—thus indicating a potential TWSS—and which tend to cluster in more decorous settings. The program then honed its skills on 2,000 sentences from twssstories.com, an online forum for "That's what she said" jokes; more practice came courtesy of fmylife.com, textsfromlastnight.com, and wikiquotes.

Comments

Popular posts from this blog

Completion