How well do hate speech, toxicity, abusive and offensive language classification models generalize across datasets?

作者：

Highlights：

• Cross-dataset model generalization for abusive language.

• Generalization of BERT, ALBERT and fastText models with respect to abusive language datasets.

• Experiments covering nine widely used public abusive speech datasets.

• Prediction of generalization by applying a random forest model.

摘要

•Cross-dataset model generalization for abusive language.•Generalization of BERT, ALBERT and fastText models with respect to abusive language datasets.•Experiments covering nine widely used public abusive speech datasets.•Prediction of generalization by applying a random forest model.

论文关键词：00-01,99-00,Hate speech,Offensive language,Classification,Generalization

论文评审过程：Received 9 August 2020, Revised 11 November 2020, Accepted 20 January 2021, Available online 9 February 2021, Version of Record 9 February 2021.

论文官网地址：https://doi.org/10.1016/j.ipm.2021.102524