Recognition of ironic sentences in twitter using attention-based LSTM

Abstract

Analyzing written language is an interesting topic that has been studied by many disciplines. Recently, due to the explosive growth of Internet, social media has become an attractive source of searching and getting information for research purposes on written communication. It is true that different words in a sentence serve different purposes of conveying the meaning while they are of different significance. Therefore, this paper is going to employ the attention mechanism to find out the relative contribution or significance of every word in the sentence. In this work, we address the problem of detecting whether a tweet is ironic or not by using Attention-Based Long Short-Term Memory Network. The results show that the proposed method achieves competitive performance on average recall and F1 score compared to the state-of-the-art results.