The importance of precision in humour classification

Abstract

Humour classification is one of the most interesting and difficult tasks in text classification. Humour is subjective by nature, yet humans are able to promptly define their preferences. Nowadays people often search for humour as a relaxing proxy to overcome stressful and demanding situations, having little or no time to search contents for such activities. Hence, we propose to aid the definition of personal models that allow the user to access humour with more confidence on the precision of his preferences. In this paper we focus on a Support Vector Machine (SVM) active learning strategy that uses specific most informative examples to improve baseline performance. Experiments were carried out using the widely available Jester jokes dataset, with encouraging results on the proposed framework. © 2011 Springer-Verlag.