Robo-comedians

To aid the next step in the robot takeover of humankind, naive researchers are letting these cyberbrains get a sense of humour. Yep, when the Terminator really does come into existence and start enslaving us all at least it will have the electro-nous to give witty Arnie-Schwarzenegger style one-liners.

The New Scientist reports on research by Julia Taylor and Lawrence Mazlack of the University of Cincinnati to program a bot to understand that most refined techniques of humour; the pun. Soon your computer might chortle out loud when it spies on your emails and finds that, for instance, they've brought out a portable stereo which looks like a big chocolate cake, which is, inevitably, called a gateaux blaster.

Given "funny" or even "trying to be funny but failing" is a hard concept to force into ones and zeroes, they approached the task at hand in the following manner. Firstly they fed it a dictionary. Yum. Then it was programmed with "knowledge" regarding how certain words relate to each other and their usual contextual and semantic usage. Add into the binary mix a pronunciation guide for good measure.

Now it's kind of easy. Feed the computer a nice potentially pun-ridden sentence. It searches its dictionary and context-relation guides to identify words that seems out of place. If it then, via knowledge of verbal pronunciation, can find a word that sounds similar but fits in better with the rest of the sentence it can deem it to be a pun and electronically HA. HA. HA. HA. at you.

To quote the Telegraph:

For example the software should understand a joke about a man asking his friend: "How was your trip to Helsinki?", to which she replies: "Terrible, all our luggage vanished into Finn Air", because "Finn" does not fit within the context of luggage and air, but does sound like "thin".

This only, so far, works on some jokes. The following was found unfunny by humourbot 1.0

Patient: "Doctor, doctor, I swallowed a bone."
Doctor: "Are you choking?"
Patient: "No, I really did!"

and, to some extent, the Poorhouse. The computer's problem was that the pun is on the word "choking" but that sort of word is semantically coherent to be used in the context of visiting the doctors. Nothing seems out of place to the novice microchip comedian, therefore error 404: humour not found.

Presumably - although undiscussed in the linked articles - the computer could, when fine tuned, start coming out with its own puns using the reverse sort of process. The day may come when all the puns possible within the English language have been discovered and recorded, although probably not before the true and exact value of pi is discovered one imagines (just kidding, maths geeks).

The New Scientist goes on to report another venture along robo-funny lines by Mihalcea et al. This seems simpler. Rather than try and establish exactly why something would be funny, it scans through a mega encyclopedia's worth of jokes to identify what sort of words, and the frequency of them, are used in them. Should these word distributions crop up again in future, they are likely to be a joke. Says Mihalcea - reporting on her findings rather than the Poorhouse's usual attitude to his employment believe it or not - "We got a lot of 'can't', 'don't', 'drunk' and 'poor'".

"Knock knock" might also be an indicator of humour in the Poorhouse's opinion. Typically, either you're about to hear something along the lines of "dunnop", "dunnop who?", "haha better clean it up then", or you're in the middle of a Guns 'n' Roses song (below), either of which some might consider humourous.


powered by ODEO


Comments

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote> <del>
  • Lines and paragraphs break automatically.
  • You may post code using <code>...</code> (generic) or <?php ... ?> (highlighted PHP) tags.
  • You may use [acidfree:xx] tags to display acidfree videos or images inline.
  • Images can be added to this post.

More information about formatting options

Captcha
This question is for testing whether you are a human visitor and to prevent automated spam submissions.
2 + 11 =
Solve this simple math problem and enter the result. E.g. for 1+3, enter 4.