Characterizing Humour: An Exploration of Features in Humorous Texts Page: 1
The following text was automatically extracted from the image on this page using optical character recognition software:
Characterizing Humour: An Exploration of Features
in Humorous Texts
Rada Mihalcea1,2, Stephen Pulman2
1 Computer Science Department, University of North Texas
2 Computational Linguistics Group, Oxford University
Abstract. This paper investigates the problem of automatic humour recognition,
and provides and in-depth analysis of two of the most frequently observed fea-
tures of humorous text: human-centeredness and negative polarity. Through ex-
periments performed on two collections of humorous texts, we show that these
properties of verbal humour are consistent across different data sets.
This paper addresses two research questions concerned with the characteristics of tex-
tual humour. First, are humorous and serious texts separable, and does this property
hold for different datasets? To answer this question, we use two different data sets of
verbal humour - a collection of short one-liners and a set of humorous news articles -
and attempt to automatically separate them from their non-humorous counterparts.
Second, if humorous and serious texts are separable, what are the distinctive features
of humour, and do they hold across datasets? In answer to this second question, we
attempt to identify some of the most salient features of verbal humour, and analyse
their occurrence in the two data sets.
While these are interesting issues in themselves, there is also a medium-term prac-
tical application for 'humour' recognition in the design of conversational agents of var-
ious types: detecting and responding appropriately to humour is a characteristic of nat-
ural human interaction that is conspicuously lacking in implemented systems. In the
longer term, by gaining insight into the mechanisms underlying humour, we hope to in-
crease our understanding of aspects of the creative use of language, i.e. uses of language
which go beyond 'banal humorless prose' and display some reflective and self-aware
properties. While these are pre-eminently displayed in creative works like novels or
poetry, they are also present in more everyday phenomena like humour.
The paper is organized as follows. We first review related work in computational
humour, and briefly cover some of the most recent methods for humour generation and
recognition. We then describe the two data sets used in this paper, and briefly overview
two machine learning techniques for text classification. Next, we address the first ques-
tion, and present the results obtained in the automatic classification of humorous and
non-humorous data sets. We then present some of the characteristics of verbal humour
as observed in an analysis of humorous texts, and provide a detailed analysis of two
of the most dominant features: human-centeredness and negative polarity. Finally, we
conclude with a discussion.
Here’s what’s next.
This paper can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Paper.
Mihalcea, Rada, 1974- & Pulman, Stephen. Characterizing Humour: An Exploration of Features in Humorous Texts, paper, February 2007; (digital.library.unt.edu/ark:/67531/metadc30988/m1/1/: accessed July 28, 2017), University of North Texas Libraries, Digital Library, digital.library.unt.edu; crediting UNT College of Engineering.