Hate Speech Detection Using Cross-Platform Social Media Data In English and German Language
Gautam Kishore Shahi, Tim A. Majchrzak
TL;DR
This study tackles the challenge of generalizing hate speech detection across platforms and languages by focusing on bilingual English–German YouTube comments and augmenting training data with cross-platform datasets from Twitter, Gab, and Wikipedia. It introduces three similarity measures—definition, content, and hate word similarity—to select helpful external data and employs a DistilBERT-based classifier, achieving a best F1 of approximately 0.74 for English and 0.68 for German when integrating cross-platform data. A dedicated annotation protocol yields a sizable, language-diverse corpus (1,892 English and 6,060 German YouTube comments) with high intercoder reliability (Cohen’s κ ≈ 0.86). The findings demonstrate that incorporating linguistically and content-similar external data can meaningfully boost performance, informing the design of more robust multilingual hate speech detectors and offering a replication package for public use.
Abstract
Hate speech has grown into a pervasive phenomenon, intensifying during times of crisis, elections, and social unrest. Multiple approaches have been developed to detect hate speech using artificial intelligence, but a generalized model is yet unaccomplished. The challenge for hate speech detection as text classification is the cost of obtaining high-quality training data. This study focuses on detecting bilingual hate speech in YouTube comments and measuring the impact of using additional data from other platforms in the performance of the classification model. We examine the value of additional training datasets from cross-platforms for improving the performance of classification models. We also included factors such as content similarity, definition similarity, and common hate words to measure the impact of datasets on performance. Our findings show that adding more similar datasets based on content similarity, hate words, and definitions improves the performance of classification models. The best performance was obtained by combining datasets from YouTube comments, Twitter, and Gab with an F1-score of 0.74 and 0.68 for English and German YouTube comments.