Cyberbullying is the use of information technology networks by individuals' to humiliate, tease, embarrass, taunt, defame and disparage a target without any face-to-face contact. Social media is the "virtual playground" used by bullies with the upsurge of social networking sites such as Facebook, Instagram, YouTube, Twitter etc. It is critical to implement models and systems for automatic detection and resolution of bullying content available online as the ramifications can lead to a societal epidemic. This paper presents a deep neural model for cyberbullying detection in three different modalities of social data, namely, textual, visual and info-graphic (text embedded along with an image). The all-in-one architecture, CapsNet-ConvNet, consists of a Capsule network (CapsNet) deep neural network with dynamic routing for predicting the textual bullying content and a convolution neural network (ConvNet) for predicting the visual bullying content. The info-graphic content is discretized by separating text from the image using Google Lens of Google Photos App. The perceptron based decision-level late fusion strategy for multimodal learning is used to dynamically combine the predictions of discrete modalities and output the final category as bullying or non-bullying type. Experimental evaluation is done on a mix-modal dataset which contains 10000 comments and posts scrapped from YouTube, Instagram and Twitter. The proposed model achieves a superlative performance with the AUC-ROC of 0.98.

No abstract

No abstract

As a constructive mode of information sharing, collaboration and communication, social media platforms offer users with limitless opportunities. The same hypermedia can be transposed into a synthetic and toxic milieu that provides an anonymous, destructive pedestal for online bullying and harassment. Automatic cyberbullying detection on social media using synthetic or real-world datasets is one of a proverbial natural language processing problem. Analyzing a given text requires capturing the existent semantics, syntactic and spatial relationships. Learning representative features automatically using deep learning models efficiently captures the contextual semantics and word order arrangement to build robust and superlative predictive models. This work puts forward a hybrid model, Bi-GRU-Attention-CapsNet (Bi-GAC), that benefits by learning sequential semantic representations and spatial location information using a Bi-GRU with self-attention followed by CapsNet for cyberbullying detection in the textual content of social media. The proposed Bi-GAC model is evaluated for performance using F1-score and ROC-AUC curve as metrics. The results show a superior performance to the existing techniques on the benchmark Formspring.me and MySpace datasets. In comparison to the conventional models, an improvement of nearly 9% and 3% in F-score is observed for MySpace and Formspring.me dataset respectively.

No abstract