Survey of Machine Learning and Deep Learning Approaches for Automated Hate Speech Detection and Sentiment Analysis in Multilingual Contexts

Abstract

The exponential rise in online hate speech has seri- ously threatened the development of an inclusive online environ- ment. The presented survey paper is an all-inclusive literature review of machine learning and deep learning techniques on automatic hate speech detection and sentiment analysis within advancements on recent developments.In that regard, several algorithms as well as architectures have been considered such as fuzzy-based convolutional neural network, ensemble methods com- bining Bi-LSTM with Naive Bayes and Support Vector Machines (SVM), object detection models like YOLO and SSD MobileNetV2 transformer-based models, including BERT. This paper will at- tempt to analyze strength and weaknesses in the identification and classification of hate speech and sentiment content within online text when such models are applied to multilingual contexts, as well as instances of code-mixing. Techniques used in feature selection in hate speech detection will be further analyzed to show which ones influence the general performance of the model.