Detecting offensive language using Chaotic Ant Lion optimization-based Ghost net in social media.

The article focuses on a novel method for detecting offensive language on social media using an Offensive Language Classification system based on the Chaotic Antlion Optimization (CAO) algorithm. The approach involves pre-processing datasets with natural language processing (NLP) techniques, extracting statistical, synthetic, and lexicon-based features, and selecting relevant features via the CAO algorithm. A Ghost network—a convolutional neural network variant—is then employed to classify text into four categories: offensive, non-offensive, swear, and offensive but not offensive. Evaluated on the SOLID and OLID datasets, the method achieved high accuracy rates of 99.27% and 98.99%, respectively, outperforming existing models such as CNN, Simple Logistic, and deep recurrent neural networks in terms of precision, recall, specificity, and computational efficiency. The study highlights the method’s potential for improved offensive language detection and suggests future integration with other swarm intelligence algorithms and adaptation for multilingual contexts.