Abstract

The governance of hate-related online communication increasingly relies on artificial intelligence, yet the scientific landscape linking computational detection methods with regulatory and ethical frameworks remains fragmented. This study provides a systematic bibliometric analysis of scholarship on the application of artificial intelligence to hate speech detection in digital environments, focusing on literature that examines hate speech and online hate through artificial intelligence, machine learning, deep learning, natural language processing, and other automated detection techniques. A dataset of 2137 publications indexed in Scopus between 2013 and 2026 was constructed and analyzed using the bibliometrix package in R. Descriptive indicators, thematic mapping, keyword co-occurrence analysis, citation structures, and temporal trend analysis were employed to examine the field’s conceptual organization, methodological evolution, and publication dynamics. The results reveal rapid annual growth, strong interdisciplinary collaboration, and a research structure dominated by language-processing methodologies, with natural language processing, machine learning, and deep learning constituting the central analytical infrastructure. Temporal patterns indicate a progression from dataset construction and feature engineering toward neural architectures, transformer models, and increasing attention to multilingual challenges. Overall, the findings indicate that the field remains predominantly oriented toward the development of scalable computational detection systems, while governance-related concerns, such as transparency, accountability, and linguistic inclusivity, emerge as structurally secondary, albeit increasingly salient, dimensions within the broader AI-centered research landscape.

IPC Classification

Keywords