Serious business makes people laugh

By BBC News Online internet reporter Mark Ward

US scientists are making people laugh in a bid to help web retailers shift more products.

We think we have found an important simplification that is faster than the method used in most collaborative filtering

Prof Ken Goldberg

The researchers have created a database of jokes and are asking people to rate them by how funny they are.

By studying comedy likes and dislikes, the scientists hope to design software that can recommend books, videos and CDs that people are bound to enjoy.

Such recommendation systems might be a boon to online retailers as figures are released which show the growth in web shopping is slowing.

Seriously funny

Led by engineering professor Ken Goldberg, scientists at the University of California, Berkeley, are designing a computer system that tries to predict which jokes people will find amusing.

The Jester 2.0 website works by asking surfers to rate 15 jokes on a scale running from "not funny" to "very funny". If a Jester tester has heard any of the jokes before, they are asked to try to remember how much they laughed the first time around.

Once the computer has digested someone's responses, it starts recommending other jokes that they should also find funny.

The research team picked jokes because they quickly reveal someone's preferences, and people are happier rating jokes than they are anything else. It is also easy for people to work out if they laughed or not. Deciding why you liked a book is much harder.

Chuckle co-efficient

The seemingly frivolous research has a serious intent. Websites such as Amazon and Firefly try to make people buy more by recommending titles to customers.

Unfortunately, the more titles and people who rate them, the longer it takes for software to work out what you might like.

Professor Goldberg and colleagues are developing a novel system that should be quicker at generating recommendations.

"We think we have found an important simplification that is faster than the method used in most collaborative filtering," he said.

Disappointing season

Instead of calculating where one person fits with regard to other respondents, Professor Goldberg and his colleagues reduce the initial responses to a short formula that can be quickly compared with others. Reactions to new jokes can be used to modify the initial result.

The software could be useful to web retailers who are keen to boost the amount people spend in online stores. Most web retailers only make money out of customers when they become regular repeat buyers.

The Xmas shopping season has not been as big a success as many retailers expected. Research by Nielsen found that many Christmas web shopping peaked in early December.

The report said: "Early indications are that this will be a disappointing season for some retailers."