Check out this article on how Natural Language Processing (NLP) and machine learning techniques can be employed to detect and classify hate speech on social media platforms like Twitter.
The authors gathered a dataset of over 24,000 tweets, categorising each tweet as containing hate speech, offensive language, or neither. Using machine learning algorithms such as Linear Regression, K-Nearest Neighbour (KNN), Random Forest, Naïve Bayes, and Decision Tree, the study aimed to find the best-performing model for identifying hate speech. After training and testing the dataset, Linear Regression emerged as the most accurate algorithm, achieving a 94% accuracy rate. KNN and Random Forest also performed well, both reaching an accuracy of 93%, while the Naïve Bayes algorithm lagged behind with 45%.
The results indicate that NLP techniques, combined with advanced machine learning models, can significantly improve the detection of abusive and hateful language online. However, some misclassifications were observed, especially in distinguishing between offensive language and hate speech, highlighting the complexities in recognising subtle differences between them.
This research underscores the potential of machine learning to enhance content moderation on social media platforms. By improving the accuracy of hate speech detection, platforms can better address harmful online behaviour, reducing the spread of hate speech and creating safer digital environments. Future research should focus on refining these models and exploring different types of hate speech to further improve content moderation efforts.