Waste-sorting policies require not only public participation but also effective communication and implementation support. This study examines public discourse surrounding Jakarta's waste-sorting policy through 413 YouTube commentsby integratingIndoBERTand SentiStrength for sentiment analysis, Liu's framework for opinion-expression analysis, LDA for topic modeling, and human annotation for model validation.Negative sentiment dominated the discourse, with IndoBERT classifying 62.0%of comments as negative and demonstrating stronger agreement with human coders than SentiStrength. Most opinions were expressed through regular-direct structures and explicit expressions. LDA identified three major discussion topics: Waste Sorting Infrastructure and Collection Practices (37.8%), Government Performance and Policy Accountability (36.1%), and Waste Disposal Behavior and Environmental Awareness(26.2%). Negative sentiment was concentrated in discussions of infrastructure and collection practices (65.4%) and government performance and accountability (63.8%). By integrating transformer-based and lexicon-based sentiment analysis with opinion-expression analysis, topic modeling, and human validation, this study provides a more comprehensive understanding of public discourse than sentiment analysis alone.The findings indicate that negative sentiment reflects not only opposition to waste sorting but also concerns about institutional and operational barriers, highlighting the need for two-way policy communication that addresses citizens' feedback and implementation challenges.