Detection of Cyberbullying in Social Media Texts using Explainable Artificial Intelligence

Loading...
Thumbnail Image

Authors

Islam, Mohammad Rafsun

Date

Type

thesis

Language

eng

Keyword

Cyberbullying , Explainable Artificial Intelligence

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

The widespread use of social media has opened the door to new forms of harassment and abuse, such as cyberbullying, that have a serious impact on individuals’ psychological health, especially children and teenagers. Therefore, research communities have recently paid attention to developing detection approaches using Natural Language Processing (NLP) combined with machine learning algorithms to identify instances of cyberbullying in social media texts such as comments, posts, and messages. Those approaches have successfully classified the social media text as either cyberbullying or non-cyberbullying. However, they are unable to determine the type of cyberbullying and the reasons why victims may be targeted based on certain characteristics. The aim of this thesis is to develop a novel detection approach that can identify the type of cyberbullying based on characteristics such as gender, religion, age, and ethnicity. This thesis has accomplished this objective by utilizing an Explainable Artificial Intelligence (XAI) technology called Local Interpretable Model-agnostic Explanations (LIME) to justify and explain the classification of text as cyberbullying. LIME enables machine learning models to capture and highlight the most influential words that affect the decision to classify a text as cyberbullying. Those influential words are utilized to re-label and update the training data. The machine learning models are then re-trained using the updated data. To evaluate the performance of the proposed approach, a simulation experiment has been conducted on a large dataset containing comments and posts from Twitter. Simulation results show that: 1) LIME provides reliable and convincing justifications and explanations for classifying a text as cyberbullying; 2) LIME enables machine learning to identify the type of cyberbullying based on characteristics such as gender, religion, age, and ethnicity; and 3) LIME improves the performance of the machine learning models in terms of classification accuracy.

Description

Citation

Publisher

License

Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Attribution 3.0 United States

Journal

Volume

Issue

PubMed ID

External DOI

ISSN

EISSN