Detail Cantuman

Social media is an online platform that facilitates interaction and expression of opinions without limitations of time and distance. However, this ease also brings up the issue of hate speech. Hate speech can take the form of discriminatory writings or conversations that incite negative actions against individuals or certain groups. To detect hate speech in social media comments, particularly on platform ‘X,’ a text classification process is conducted. Text classification is a data mining technique that categorizes text based on existing patterns. The main challenge in sentiment analysis lies in processing unstructured text data, such as slang, abbreviations, and mixed languages, which require contextual representation approaches. In this study, the SVM method is used to classify hate speech and is combined with CountVectorizer feature extraction on social media comments. Among the four SVM kernels, the race category achieved the highest accuracy in each kernel: linear kernel 96%, polynomial kernel 72%, RBF kernel 95%, and sigmoid kernel 97%. Meanwhile, the lowest accuracy in each kernel was found in the other category: linear kernel 74%, polynomial kernel 54%, RBF kernel 68%, and sigmoid kernel 74%. This study emphasizes that each kernel has its advantages and disadvantages for each category of hate speech classification.