On the Generalization of Figurative Language Detection: The Case of Irony and Sarcasm

摘要

The automatic detection of figurative language, such as irony and sarcasm, is one of the most challenging tasks of Natural Language Processing (NLP). In this paper, we investigate the generalization capabilities of figurative language detection models, focusing on the case of irony and sarcasm. Firstly, we compare the most promising approaches of the state of the art. Then, we propose three different methods for reducing the generalization errors on both in- and out-domain scenarios.