NLP classifies sentences / paragraphs as funny

Question

Is there a way to classify a specific sentence / paragraph as funny. There is very little guidance on where to go next.

Answer 1

This is a study called Computational Humour . This is an interdisciplinary field that takes elements from computational linguistics, psycholinguistics, artificial intelligence, machine learning, etc. They try to find out what it is that makes funny stories or jokes (for example, an unexpected mix or use of a taboo theme in an amazing way, etc.) and apply it to the text (either create a funny story or measure the "fun" of the text) .

There are books and articles about this (e.g. Graeme Ritchie ).

Answer 2

Yes, you should use Training Corpora to create a predictive model that can detect fun sentences. This is sometimes known as "Mood Analysis" in the literature. Take a look at this article on Analyzing Moods with LingPipe .

If you can use Java, you can use their library (see the licensing matrix ). I found this very useful, not quite in the same context as you.

Answer 3

The only way to do this is to get a couple of thousand people (monkeys will not do it, sorry) to view thousands of funny sentences / stories, evaluate them, and then create some kind of expert system / neural network from it. Given the problem and its subjectivity (a thing funny to one person may not be funny, even offensive, another), I would say that this is an impossible task.

Answer 4

You can use the same method as spam filters. Instead of spam / non-spam, you classify funny / non-funny. Look at naive Bayes classifiers for more information.

http://en.wikipedia.org/wiki/Naive_Bayesian_classification

Answer 5

Also, try Computational Humour @Google Scholar if you are serious about getting into the field. Mood analysis was also mentioned, see wikipedia .

Of course, it all depends on what your scope and goals are ...