Beyond Trolling: Fine-Grained Detection of Antisocial Behavior in Social Media During the Pandemic DOI Creative Commons
Andrew Asante, Petr Hájek

Information, Год журнала: 2025, Номер 16(3), С. 173 - 173

Опубликована: Фев. 26, 2025

Antisocial behavior (ASB), including trolling and aggression, undermines constructive discourse escalates during periods of societal stress, such as the COVID-19 pandemic. This study aimed to examine ASB on social media pandemic by leveraging a novel annotated dataset state-of-the-art transformer models for detection classification categories. Specifically, this examined within gold-standard corpus tweets collected from Ghana 21-day lockdown. Each tweet was meticulously into categories or non-ASB, enabling comprehensive analysis online behaviors. We employed three transformer-based language (BERT, RoBERTa, ELECTRA) compared their performance against traditional machine learning models. The results demonstrate that approaches substantially outperformed baseline models, achieving high accuracy across both binary multiclass tasks. RoBERTa excelled in detection, attaining 95.59% an F1-score 94.99%, while BERT led classification, with 94.38% 93.92%. Trolling emerged most prevalent type, reflecting polarizing nature interactions highlights potential detecting diverse behaviors emphasizes implications crises. findings provide foundation enhancing moderation tools fostering healthier environments.

Язык: Английский

Beyond Trolling: Fine-Grained Detection of Antisocial Behavior in Social Media During the Pandemic DOI Creative Commons
Andrew Asante, Petr Hájek

Information, Год журнала: 2025, Номер 16(3), С. 173 - 173

Опубликована: Фев. 26, 2025

Antisocial behavior (ASB), including trolling and aggression, undermines constructive discourse escalates during periods of societal stress, such as the COVID-19 pandemic. This study aimed to examine ASB on social media pandemic by leveraging a novel annotated dataset state-of-the-art transformer models for detection classification categories. Specifically, this examined within gold-standard corpus tweets collected from Ghana 21-day lockdown. Each tweet was meticulously into categories or non-ASB, enabling comprehensive analysis online behaviors. We employed three transformer-based language (BERT, RoBERTa, ELECTRA) compared their performance against traditional machine learning models. The results demonstrate that approaches substantially outperformed baseline models, achieving high accuracy across both binary multiclass tasks. RoBERTa excelled in detection, attaining 95.59% an F1-score 94.99%, while BERT led classification, with 94.38% 93.92%. Trolling emerged most prevalent type, reflecting polarizing nature interactions highlights potential detecting diverse behaviors emphasizes implications crises. findings provide foundation enhancing moderation tools fostering healthier environments.

Язык: Английский

Процитировано

0