2026-05-13T16:20:38.272Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:10150 2025-06-13T13:58:39Z 3

Twitter toxic comment identification in digital media and advertising using NLP and optimized classifiers 2024 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/10150 https://www.atlantis-press.com/proceedings/iciitb-24/126002403 Ј. Гајић L. Dražeta Л. Бабић Ј. Каљевић Д. Јовановић Л. Јовановић Cyberbullying is a form of harassing, intimidating and harming other people through electronic media like social networks or messaging platforms. Typical forms of cyberbullying include messages containing harmful text, photos or videos that will embarrass the target, and excluding the individual from groups and chats. Unfortunatelly, it may lead to sincere psychological problems of the target, including disorders like depression, anxious behavior, lack of self-esteem, or even worse, suicidal thoughts and self-hurting. The research presented herein proposes a hybrid approach that includes natural language processing and machine learning XGBoost model optimized by an altered variant of Botox optimization metaheuristics for classification of toxic tweets on a real-world dataset. The experimental results have shown considerable prospect of application of machine learning models in solving this serious and important problem. bookPart Atlantis Press 171 187 10.2991/978-94-6463-482-2_12 Proceedings of the 2nd International Conference on Innovation in Information Technology and Business (ICIITB 2024), Advances in Computer Science Research 113