Detection and Fine-Grained Classification of Cyberbullying Events

Proceedings of Recent Advances in Natural Language Processing, pages 672–680, Hissar, Bulgaria, Sep 7–9 2015. Detection and Fine-Grained Classification of Cyberbullying Events Cynthia Van Hee∗, Els Lefever∗, Ben Verhoeven†, Julie Mennes∗, Bart Desmet∗ Guy De Pauw†, Walter Daelemans† and V´eronique Hoste∗ ∗ LT3 - Language and Translation Technology Team, Ghent University Groot-Brittanni¨elaan 45, 9000 Ghent, Belgium firstname.lastname@ugent.be † CLiPS - Computational Linguistics Group, University of Antwerp Prinsstraat 13, 2000 Antwerp, Belgium firstname.lastname@uantwerpen.be Abstract In the current era of online interactions, both positive and negative experiences are abundant on the Web. As in real life, neg- ative experiences can have a serious im- pact on youngsters. Recent studies have reported cybervictimization rates among teenagers that vary between 20% and 40%. In this paper, we focus on cyberbul- lying as a particular form of cybervictim- ization and explore its automatic detection and fine-grained classification. Data con- taining cyberbullying was collected from the social networking site Ask.fm. We de- veloped and applied a new scheme for cyberbullying annotation, which describes the presence and severity of cyberbully- ing, a post author’s role (harasser, victim or bystander) and a number of fine-grained categories related to cyberbullying, such as insults and threats. We present exper- imental results on the automatic detection of cyberbullying and explore the feasibil- ity of detecting the more fine-grained cy- berbullying categories in online posts. For the first task, an F-score of 55.39% is ob- tained. We observe that the detection of the fine-grained categories (e.g. threats) is more challenging, presumably due to data sparsity, and because they are often ex- pressed in a subtle and implicit way. 1 Introduction Young people are gaining more frequent and rapid access to online, mobile and networked media. Al- though most of the time, children’s Internet use is harmless, there are some risks associated with the online activity, such as the use of social network- ing sites (e.g. Facebook). The anonymity and free- dom provided by social networks makes children vulnerable to threatening situations on the Web, such as grooming by paedophiles or cyberbully- ing. According to Smith et al. (2008), cyberbul- lying is defined as an aggressive, intentional act carried out by a group or individual, using elec- tronic forms of contact, repeatedly and over time against a victim who cannot easily defend him or herself. Their definition is based on three criteria (repetitiveness, intentionality, and an imbalance of power between the harasser and the victim) that are recognized as inherent characteristics of bul- lying by Olweus (1996). Some doubt exists, nev- ertheless, as to whether all three criteria are nec- essary conditions for cyberbullying. For example Dooley and Cross (2010) and Grigg (2010) stress that online posts are persistent, a single aggres- sive act can result in continued and widespread ridicule for the victim. Furthermore, it is hard to decide upon intentionality since online communi- cation is prone to misinterpretation (Kiesler et al., 1984; Vandebosch et al., 2006). Finally, the as- sessment of a power imbalance is complicated in online bullying as it …

Content truncated to 500 words