A CORPUS-BASED ANALYSIS OF CYBERBULLYING LANGUAGE TROUGH LEXICAL COLLOCATIONS AND SEMANTIC PROSODY ON TELEGRAM

Abstract

As digital communication platforms proliferate, cyberbullying has evolved from isolated harassment into a linguistically complex, multi-turn phenomenon. This article applies corpus linguistics methodologies to analyze toxic discourse specifically within public Telegram channels. By compiling a specialized corpus of flagged user comments, this study investigates how linguistic devices – such as semantic prosody, algorithmic evasion (algospeak), and highly contextual codeswitching – are utilized to perpetuate digital harassment. The findings demonstrate that cyberbullying language on Telegram relies heavily on negative semantic prosody and multi-turn escalation, posing unique challenges for static moderation models. This research underscores the necessity of context-aware, corpus-driven approaches to detect and mitigate online abuse.