Publication Date

Fall 2021

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Teng Moh

Keywords

Multimodal system, Multi-Layer Perceptron model, cyberbully, Twitter dataset, Convolutional Neural Network (CNN), Tensor Fusion Network, VGG-19 Network.

Abstract

Cyberbullying detection is one of the trending topics of research in recent years, due to the popularity of social media and the lack of limitations about using electronic communications. Detection of cyberbullying may prevent some bullying behaviors online. This paper introduced a Multimodal system that makes use of Convolutional Neural Network (CNN), Tensor Fusion Network, VGG-19 Network, and Multi-Layer Perceptron model, for the purpose of cyberbullying detection. This system can not only analyze the messages sent but also the extra information related to the messages (meta-information) and the images contained in the messages. The proposed system was trained and tested on Twitter datasets, achieving accuracy scores of 93%, which was 4% higher than scores of the benchmark text-only model using the same dataset and 6.6% higher than previous work. With the results, we believed that the proposed system performs well and it will provide new ideas for future works.

Available for download on Thursday, January 19, 2023

Share

COinS