Capturing the Spectrum of Social Media Conflict: A Novel Multi-objective Classification Model
Keywords: Deep Learning, Social Media Hate, Social Media Conflicts, Decision Transformer Framework
TL;DR: Multi-class classification model incorporating novel class based reward functionality, aimed at tackling a range of user conflicts on social media.
Abstract: Social media has emerged as a widespread phenomenon, with numerous users engaging in observing, creating, and distributing content. The growing content has led to user conflicts, encompassing bullying, aggression, harassment, and threats. Consequently, recent research has aimed to identify and address these openly hostile forms of social conflict. However, the less overtly hostile yet equally damaging types of conflict, including teasing, criticism, and sarcasm, have been overlooked in current studies.
Our aim is to detect these subtle forms of conflict, while also including openly hostile forms, by developing a novel multi-objective classification model. This innovative approach leverages class based reward functions to improve model performance. Reward functions serve as potent signals capable of mitigating the intricacies of misclassification in multi-class scenarios. By incorporating various rewards within the model architecture, harnessing the power of a decision transformer, we achieved significant improvements in classification performance. Our experiments on three datasets demonstrate superior recall, precision, f1-score, and accuracy compared to traditional state-of-the-art deep learning classifiers. Furthermore, we analyse class ambiguity and its impact on model performance as well as conducting thematic analysis on model misclassifications. We will share the code and datasets at github.com/anonymous.
Submission Number: 10
Loading