A research paper published on October 24, 2026, on the arXiv platform (ID 2610.30885v1) introduces the “Ontology-Driven Cyberbullying Behavior and Network Analysis Framework” (ODCBNA), which aims to systematically analyze bullying patterns in the Bengali-language online environment. Centered on the perpetrator, this framework integrates five major modules: ontology design driven by user needs, knowledge graph construction, behavioral modeling, community analysis, and impact modeling.
The study, grounded in the Bengali‑language contextualized multilingual cyberbullying corpus (BenCoDiC), identified 1,063 users who had posted at least three bullying comments or replies as perpetrators and constructed a 64‑dimensional vector of behavioral and contextual features for each. Using cosine similarity and k‑nearest neighbors modeling, a weighted network with 27,554 edges was generated; the Louvain algorithm detected six major communities (with modularity around 0.495) and 27 subcommunities.
Feature contribution analysis shows that the three dimensions—“video,” “influence,” and “user profile”—are most explanatory for behavioral similarity and community differentiation, while the contextual dimension supports result interpretability. In influence propagation analysis under community constraints, the IM Greedy algorithm yields a modest positive gain across all six communities; although different seed-selection methods exhibit low initial seed overlap, their diffusion coverage is comparable.