Analisis Framing dan Sentimen Media Online Indonesia dalam Pemberitaan Tom Lembong

Abstract

This study analyzes the framing and sentiment of Indonesian online media coverage of the sugar import corruption case involving the former Secretary-General of the Ministry of Trade, Tom Lembong, during the period from late 2024 to mid-2025, focusing on three major news outlets: Detik, Kompas, and Tempo. Framing analysis applies Robert M. Entman’s model, operationalized through the Gemma 3:4b language model using a zero-shot approach, while sentiment analysis is conducted using the IndoRoBERTa model. The results demonstrate high model reliability, indicated by a Mean Confidence of 0.92, Low Confidence of 4.6%, and a Sentiment Entropy of 0.90 bits, with sentiment distribution dominated by the Neutral class (74.05%). Framing consistency is validated through semantic-based metrics, yielding an Intra-frame Similarity of 0.41, an Inter-frame Similarity of 0.68, and a Frame Separability Index of 0.60. Statistical association tests reveal a significant yet moderate relationship between framing and sentiment (χ² = 119,915; p < 0.05; Cramer’s V = 0.33), indicating that framing influences sentiment formation without fully determining the emotional tone of news coverage. Overall, the three online media outlets tend to present the case in a factual and neutral manner, with negative sentiment primarily concentrated in the Moral Judgment frame, underscoring the effectiveness of a large language model–assisted framing analysis for examining how Indonesian media construct narratives around high-profile corruption cases.