Data Expansion Using WordNet-based Semantic Expansion and Word Disambiguation for Cyberbullying Detection

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1761–1770 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 1761 Data Expansion Using WordNet-based Semantic Expansion and Word Disambiguation for Cyberbullying Detection Md Saroar Jahan⋆, Djamila Romaissa Beddiar⋆, Mourad Oussalah⋆, Muhidin Mohamed⋄ ⋆University of Oulu, CMVS, BP 4500, 90014, Finland ⋄ Operations and Information Management, Aston University, B4 7ET, UK {Md.Jahan, Mourad.Oussalah, Djamila.Beddiar}@oulu.fi, m.mohamed10@aston.ac.uk Abstract Automatic identification of cyberbullying from textual content is known to be a challenging task. The challenges arise from the inherent structure of cyberbullying and the lack of labeled large-scale corpus, enabling efficient machine-learning-based tools including neural networks. This paper advocates a data augmentation-based approach that could enhance the automatic detection of cyberbullying in social media texts. We use both word sense disambiguation and synonymy relation in WordNet lexical database to generate coherent equivalent utterances of cyberbullying input data. The disambiguation and semantic expansion are intended to overcome the inherent limitations of social media posts, such as an abundance of unstructured constructs and limited semantic content. Besides, to test the feasibility, a novel protocol has been employed to collect cyberbullying traces data from AskFm forum, where about a 10K-size dataset has been manually labeled. Next, the problem of cyberbullying identification is viewed as a binary classification problem using an elaborated data augmentation strategy and an appropriate classifier. For the latter, a Convolutional Neural Network (CNN) architecture with FastText and BERT was put forward, whose results were compared against commonly employed Na¨ıve Bayes (NB) and Logistic Regression (LR) classifiers with and without data augmentation. The research outcomes were promising and yielded almost 98.4% of classifier accuracy, an improvement of more than 4% over baseline results. Keywords: cyberbullying detection, social media, disambiguation, dataset expansion, WordNet 1. INTRODUCTION The multiplication of cases associated with hate speech (HS) and cyberbullying nowadays, especially on social me- dia platforms (Gong et al., 2014), raises concerns about the efficiency of adopted measures. This becomes a challenge for policymakers, educators, sociologists, civil society ac- tors, and researchers. Indeed, the advances in Web 2.0 technology, which led to the emergence of anonymously user-generated content, have witnessed the proliferation of cyberbullying cases in public forums and political dis- course. Research in cyberbullying has been mainly multi- disciplinary by nature, where several definitions have been promoted in the literature depending on the research focus. Possibly, the definition that best accommodates our work in this paper is that reported in (Rosa et al., 2019) where cy- berbullying is associated with an individual’s or group’s re- peated intentional aggressive behavior towards other peers with the intent of harming them by sending offensive con- tent or engaging in other forms of social aggression through the use of digital technologies. Another related cyberbully- ing definition is provided by (Chen, 2011) as an ‘aggressive and intentional act carried out by a group or individual, us- ing electronic forms of contact, repeatedly and over time, against a victim who can not easily defend him or …

Content truncated to 500 words