Hate Speech Detection on Twitter Using Multinomial Logistic Regression Classification Method

摘要

In today's social media, especially Twitter is very important for the success and destruction of one's image due to the many sentences of opinion that can compete the users. Examples of phrases that mean evil refer to hate speech to others. Evil perspectives can be categorized in hate speech, which hate speech is regulated in Article 28 of the ITE Law. Not a few people who intentionally and unintentionally oppose a social media that contains hate speech. Unfortunately social media does not have the ability to aggregate information about an existing conversation into a conclusion. One way to draw conclusion from aggregation results is to use text mining. In this paper to classify whether the text in the sentence contains elements of hate speech or not. The author hopes in this paper can make how to classify element of hate speech in text by computer, which later speech of the can be recognized. By using Multinomial Logistic Regression method. The author hopes after this application the computer can know and classify the existence of hate speech on a text from social media Twitter. From the results of tests that have been done the average precision of 80.02, recall 82%, and accuracy of 87.68%.