SatiricLR: a Language Resource of Satirical News Articles

SatiricLR: a Language Resource of Satirical News Articles Alice Frain and Sander Wubben Tilburg University Tilburg, Netherlands E-mail:A.Frain@tilburguniversity.edu, S.Wubben@tilburguniversity.edu Abstract In this paper we introduce the Satirical Language Resource: a dataset containing a balanced collection of satirical and non satirical news texts from various do- mains. This is the first dataset of this magnitude and scope in the domain of satire. We envision this dataset will facilitate studies on various aspects of of sat- ire in news articles. We test the viability of our data on the task of classification of satire. Keywords: language resource, text categorisation, figurative language, irony, satire 1. Introduction With the ever-growing amount of satire on web and the complex nature of Figurative Language Processing (FLP), studies in the field of satire are becoming more and more important. Whereas most NLP research centers on the modeling of natural language, FLP attempts to find elements in language that aid in the computational processing of figurative forms of language. As outlined by Reyes, Rosso and Buscaldi (2012), figurative language differs from literal language as it represents meaning by using linguistic devices such as ambiguity, irony, metaphors etcetera. Uncovering the true meaning of these devices relies on our cognitive abilities, which help us reason beyond the mere syntax of a sentence. One can therefore imagine that automatically detecting satire is a challenging NLP task. In order to get to the real meaning of a text and to detect whether it is satire, a model will need to have access to contextual knowledge and has to be able to rely on various social and cognitive capacities which are difficult to computationally represent. Despite its challenging nature, a lot of work has been done in the field of FLP with promising results. Common areas of research include: the automatic detection of similes and metaphors (Veale & Hao 2007, 2010), detecting humorous texts (Mihalcea & Pulman 2007), irony detection (Carvalho, Sarmento, Silva & Oliveira 2009; Reyes & Rosso, 2011; Reyes, Rosso & Buscaldi, 2012) and sarcasm detection (Davidov, Tsur & Rappoport, 2010). Studies related to the automatic recognition of satire are scarce. Nonetheless, satire knows various definitions in academic literature. Claridge (2011) emphasizes the critical aspect of satire and reasons that satire accomplishes its satirical goal by using various figurative devices such as exaggeration, hyperbole and irony. Notice how different figurative devices are incorporated in the aforementioned definitions. This demonstrates that even though they are different devices, they are often used in combination with each other to achieve a satirical intention. This includes related concepts such as irony and humor. We therefore argue that automatic satire detection is closely related to humor, sarcasm and irony detection. This paper builds upon the study of Burfoot & Baldwin (2009). To our knowledge, this has been the only study that focused on the detection of satire. This paper contributes to current literature by (i) using a larger dataset consisting of an equal distribution of satirical and non-satirical articles from the topical domains politics, entertainment and …

Content truncated to 500 words