Transformer based contextualization of pre-trained word embeddings for irony detection in Twitter

作者：

Highlights：

• To study the irony detection problem for the English and the Spanish languages on two widely used corpora.

• To present an approach based on Transformer Encoders for contextualizing pre-trained Twitter word embeddings.

• To propose strategies for understanding the behavior of Transformer Encoder models in irony detection problems.

• To provide, under request, the implementation of our system for research purposes.

摘要

•To study the irony detection problem for the English and the Spanish languages on two widely used corpora.•To present an approach based on Transformer Encoders for contextualizing pre-trained Twitter word embeddings.•To propose strategies for understanding the behavior of Transformer Encoder models in irony detection problems.•To provide, under request, the implementation of our system for research purposes.

论文关键词：Irony detection,Twitter,Deep learning,Transformer encoders

论文评审过程：Received 3 January 2020, Revised 28 March 2020, Accepted 4 April 2020, Available online 16 April 2020, Version of Record 16 April 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2020.102262