<![CDATA[This study examines public discourse surrounding Danantara by applying BERTopic optimized with UMAP dimensionality reduction and HDBSCAN clustering to model thematic structures within Indonesian-language tweets. The increasing volume of digital conversations and uncertainty surrounding Danantara necessitate analytical approaches that go beyond conventional sentiment classification to capture thematic diversity. The dataset underwent extensive preprocessing, including duplication removal and temporal anomaly detection, to reduce noise and mitigate distortion from non-organic conversational bursts. BERTopic was configured using IndoSBERT-large, a transformer-based embedding model specifically designed for Indonesian-language semantic representation, alongside optimized UMAP–HDBSCAN parameters to ensure stable clustering. The analysis identified fifteen distinct topics spanning economic optimism, energy and state-owned enterprises, concerns over legality and accountability, political narratives, and informal or humorous interactions. The findings demonstrate that public discourse on Danantara is highly heterogeneous and shaped by socio-political dynamics on social media. Overall, the proposed approach proves effective in uncovering layered discourse structures, as reflected in the coherence and diversity of the generated topics, and it provides a data-driven foundation for analyzing public perception and informing policy communication strategies.]]>
Copyrights © 2026
