Pemodelan Topik Artikel Berita Menggunakan Structural Topic Model dan Latent Dirichlet Allocation

Abstract

An online news portal is one of the technologies in the form of online media that provides information services in the form of news articles. The number of news articles on online news portals continues to grow over time and more news article data will be available. A large amount of data is a challenge in itself to be processed into a more useful form, namely by conducting topic modeling based on news article data so that the data can be categorized based on the topics discussed in it. Topic modeling groups text data into a specific set of topics based on their similarities. In this study, the dataset used was 44,425 news articles from November 2021 to March 2022 which were taken from the online news portal detik.com. News exploration was carried out by topic modeling using two methods, Latent Dirichlet Allocation (LDA) and Structural Topic Model (STM). The LDA method produces 8 topics derived from the calculation of the highest probabilistic coherence value. The STM method produces 11 topics based on the highest semantic coherence and exclusivity values.