Topic Modeling and Sentiment Analysis of YouTube Podcast “Susahnya Jadi Perempuan” Using LDA and SVM Algorithms

Abstract

Youtube Podcast “Susahnya Jadi Perempuan” which addresses feminist issues has garnered attention from viewers of Najwa Shihab Channel. In this digital era, sentiment analysis of audience response is needed to understanding public perception. Method that can used to determine discussion topics and analyze sentiments is using Latent Dirichlet Allocation and Support Vector Machine. Analysis of 10.979 comments using LDA identified two subtopics: discussion about the invented speakers and gender roles in daily life. Along with this, sentiment analysis using an optimized SVM (C=1, gamma=1, kernel=linear) which classified sentiments into Positive, Negative, and Neutral categories with an accuracy of 67%. The main challenge was the low recall value for Neutral sentiments classification. The results showed that in subtopic 0, there were 3.503 Negative sentiments, 3.255 Positive sentiments, and 822 Neutral sentiments. In subtopic 1, there were 1.485 Negative sentiments, 1.671 Positive sentiments, and 243 Neutral sentiments.