Sarcasm Detection Using RNN with Relation Vector
Satoshi Hiai, Kazutaka Shimada · International Journal of Data Warehousing and Mining · 2019
Sarcasm detection has been treated as a task that classifies text as sarcastic or non-sarcastic. Sarcasm detection is a significant challenge for sentiment analysis because sarcasm involves a positive expression with a negative meaning. Surface information in text is commonly used as a classification feature. However, the authors must consider both surface and non-surface features. In this article, the authors focus on relation information between pairs of role expressions, such as “boss and staff,” and propose a sarcasm detection method based on surface and relation information. First, the authors extract role pairs from a corpus. Then, the authors construct a relation vector generated from these role pairs and incorporate the relation vector into a recurrent neural network model. The authors evaluated the proposed method by comparing it to previously proposed methods. The results demonstrate the effectiveness of introducing the relation vector to sarcasm detection.