Path: utzoo!utgpu!watmath!clyde!att!gargoyle!hajek
From: hajek@gargoyle.uchicago.edu (Gregory Michael Hajek)
Newsgroups: comp.ai
Subject: Re: Understanding Jokes
Message-ID: <258@gargoyle.uchicago.edu>
Date: 13 Nov 88 09:11:44 GMT
References: <3244@sdsu.UUCP> <333@igor.Rational.COM> <777@wsccs.UUCP>
Reply-To: hajek@gargoyle.uchicago.edu.UUCP (Greg Hajek)
Organization: U. Chicago Computer Science Dept.
Lines: 21

In article <777@wsccs.UUCP> dharvey@wsccs.UUCP (David Harvey) writes:
>> In article <3244@sdsu.UUCP>, caasi@sdsu (Richard Caasi) writes:
>>>
>>>Does anyone know of any work or program that takes a joke as
>>>input and outputs some number, say between 0 to 10, based on
>>>the funniness of the joke?
>
>Actually
>though, there must be some other measure that could be made into a
>rule based system..... (any takers?).

Well, for a first stab, puns aren't too tough; a reasonable measure could be the
amount of backtracking a pun requires the hearer to make to identify the
actual meaning of a pun, as opposed to its initially conveyed meaning.  .  It 
was all the money I had."  The conveyed meaning is that "bucks" = "male deer",
when the actual meaning is "money".  Of course, by this measure garden path
sentences are hysterical, but you could always put a limit on how much
backtracking is just too much work (it becomes too annoying to enjoy the joke).
----------
Greg Hajek {....!uwvax!oddjob!tank!gargoyle!hajek}
"She was the only woman I had ever met who said 'thanks' with an 'x'."