A Parallel Dual-Channel Chinese Offensive Language Detection Method Combining BERT and CTM Topic Information

With the development of intelligent technology, the application of detection models in various fields becomes more and more important. In this study, a novel detection model (BCOLD) is developed, which is not only suitable for language detection, but also can be widely used in fields such as medical text and image identification.The BCOLD model first utilizes BERT-generated word vectors to capture contextual details, and then combines them with CTM-generated topic vectors to understand the core themes of the text. This fusion strategy enhances the model's detection capability and understanding of the deeper meaning of the text. The fused vectors are fed into DPCNN and TextCNN models in parallel to capture complex semantic structures and local features, and the feature representation is further optimized by the Multi-Head Attention mechanism. The experimental results show that the BCOLD model performs well in language detection, provides an efficient and accurate solution for automatic detection and classification, and exhibits a wide range of application prospects.