2026-05-04T22:06:07.009Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:2:11468 2025-06-24T13:59:17Z 2

Detecting Harassment in User Comments: Two-Tier Machine Learning and Metaheuristics Approach with Natural Language Processing 2025 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/2/11468 https://link.springer.com/article/10.1007/s42979-025-04122-x M. Todorovic S. Kozakijevic L. Jovanovic L. Babic M. Antonijevic V. Zeljkovic M. Zivkovic N. Bacanin As a pillar of modern society the internet has revolutionized communication, offering instantaneous ways to write to, call, or video call people across the world at a low cost. Thus, social circles expand, and the Internet allows strangers to interact easily. Although this has many positive outcomes, it also magnifies the opportunities for harassment, especially coupled with the option of anonymity. Harassment, be it online or in person, can have severe negative consequences on the mental health of individuals exposed to it. Detection and adequate moderation of such interactions can pose a challenge that artificial intelligence (AI) can aid in tackling. Using the natural language processing (NLP) approach, AI can analyze textual data and learn to recognize and flag problematic communication. This work utilizes a two-layer approach for feature selection and reduction, in addition to convolutional neural networks (CNN) and AdaBoost classifier optimized by a modified genetic algorithm (GA) based classification, for the identification of aggressive comments from YouTube videos will be demonstrated. With an accuracy of 0.993815, the hybrid model CNN-AB-SAGA, optimized using this updated approach, shows favorable results. The low amount of estimators it has also implies a reduced likelihood for overfitting in noisy text based data sets. This makes CNN-AB-SAGA relevant and suggests its potential for successful harassment identification. article 10.1007/s42979-025-04122-x 6 573 2661-8907 SN Computer Science