TASR: Adversarial learning of topic-agnostic stylometric representations for informed crisis response through social media

作者：

Highlights：

• Our generalized approach does not depend on manual labeling of topics or domains upon the emerging of a crisis event.

• A new stylometric-based design serves as an alternate and improved solution for explainable crisis information filtering.

• A novel unsupervised adversarial learning approach removes the potential topic biases from the stylometric representations.

摘要

•Our generalized approach does not depend on manual labeling of topics or domains upon the emerging of a crisis event.•A new stylometric-based design serves as an alternate and improved solution for explainable crisis information filtering.•A novel unsupervised adversarial learning approach removes the potential topic biases from the stylometric representations.

论文关键词：Crisis informatics,Adversarial learning,Stylometric representation,Explainability

论文评审过程：Received 14 October 2021, Revised 30 November 2021, Accepted 26 December 2021, Available online 28 January 2022, Version of Record 28 January 2022.

论文官网地址：https://doi.org/10.1016/j.ipm.2021.102857