Lexicon-enhanced hate speech detection on Vietnamese social network data

Abstract

This paper applies two lexicon enhancement methods, that are lexica feature extraction and retrofitting to improve the accuracy of hate speech detection problem on Vietnamese social network data. The experiments were conducted on multiple datasets to achieve the statistical significance of the experimental results. The results show that the use of retrofitting lexicon enhancement improves the accuracy of hate speech detection. This paper also introduces a dictionary consisting of hateful words that can be used for lexicon enhancement for hate speech detection on Vietnamese social network data.