Satirical News Headline Detection Based on BERT+LSTM
Yang Chen, Chengqiong Ye · 2024
It is an important task to correctly identify the satirical features of news. Sarcasm detection, as a task of text classification in natural language processing, requires deep semantic information. Under the framework of Keras and Tensorflow, this experiment preprocessed text data, obtained word vector through BERT model, and then input LSTM model to obtain feature vector, and finally classified after dimensionality reduction. The experimental results on the standard satire detection dataset show that the proposed model has a high accuracy of irony detection classification.