Sentiment and Emotion Analysis of Climate Related Social Media Content

This study proposes a domain-specific sentiment and emotion analysis framework to address the limitations of general tools in capturing nuanced environmental discourse on social media. Using a hybrid model combining Pointwise Mutual Information (PMI) and Logistic Regression, along with NRCLex for emotion detection and BERTopic for topic modeling, the research analyses social media posts from 2014 to 2023. The findings show a predominance of negative sentiment, particularly on Twitter and Facebook, while Instagram displays more positivity. Common topics include climate change, plastic pollution, and air quality. The proposed model demonstrates improved accuracy over existing tools and is validated through expert annotations, offering a robust, multi-platform approach to understanding public perceptions of environmental issues— beneficial for policymakers, researchers, and communicators