OFFENSIVE LANGUAGE DETECTION BASED ON SEMANTIC SPELLING COMPREHENSION AND GATED ATTENTION MECHANISM

How to automatically detect offensive language information spread on the Internet is one of the hot research contents in the field of natural language processing. Aiming at the characteristics of semantic expression and spelling habits in offensive language, this paper proposes a offensive language detection method based on semantic spelling understanding and gating attention mechanism. This method used a self-attention mechanism to obtain the semantic features of the text, used a convolutional neural network to extract the spelling features of the text, and used a combination of early feature fusion and gated attention mechanism to fuse semantic and spelling features. Experimental results on two public datasets show that the proposed model can effectively extract the semantic features of offensive language and improve the performance of offensive language detection.