Computational Topic Discovery and Content Analysis of Digital News Media in Indonesia Using a Structural Topic Modeling Approach on Q1 2026 Nusantara TV Data

Abstract

The swift expansion of digital news platforms has produced substantial volumes of unstructured textual data, requiring sophisticated computational techniques for content surveillance and public discourse examination. This study examines the thematic landscape of Nusantara TV, a leading Indonesian digital news platform, in the first quarter (Q1) of 2026. We use a Structural Topic Model (STM) on a dataset of 7,163 video titles and descriptions that we got from the YouTube Data API to find hidden themes and how common it tend to be. We use a multi-stage preprocessing pipeline that includes custom Indonesian stopword filtering and spectral initialization to make sure the model is stable. The STM finds seven different thematic clusters, from national economic performance and government social programs to global geopolitical conflicts and regional social issues. Quantitative results, examined via anticipated topic proportions and word probability indices, indicate that economic recovery and government-driven social initiatives predominated the news cycle during this timeframe. This study shows that automated topic modeling works well for giving media researchers and policymakers a high-level understanding of what's going on, and it also sets up a framework for long-term studies of digital journalism in Southeast Asia.