Chinese Sarcasm Detection Using Title-Attention and Bidirectional LSTM Base on Bert

Sarcasm is a special linguistic phenomenon, and sarcasm recognition is a challenging natural language processing task. The complexity of Chinese grammar means that sarcasm recognition in Chinese text will be more difficult. Sarcasm often relies on context, implicit language structures, and complex semantic relationships, which makes traditional text classification and sentiment analysis methods difficult to deal with. This article provides a deep learning model that can deeply mine the emotions contained in the text and combine the dependencies between sentences to identify the irony contained in the sentences and analyze the true emotions expressed by the sentences. Moreover, this article proposes a Title-Attention so that the input comment text can be well combined with the news title text. Experiments have proven that this model can help identify the true emotional tendencies in texts and mine sarcastic remarks.