PlastiNews: An In-Depth Analysis of Media Coverage Regarding Plastic Pollution in India and the United States Using a Large Language Model (LLM)
Keywords: Large Language Models, Plastic Pollution, News Analysis
TL;DR: We used a large language model to analyze news about plastic pollution from India and the United States, discovering patterns in coverage across the two countries that could be highly useful for government and activist organizations.
Abstract: A 2023 United Nations Environment Programme report found that 3200 out of the 7000 substances associated with plastic have hazardous properties. Plastic pollution is clearly a global health crisis. The issue is also multi-faceted: the International Union for the Conservation of Nature calls it an issue that has “multiple sources and actors.” Despite this, media coverage of plastic pollution has largely focused on only a few of its forms. This can create a dangerous restriction in public perspective, which plays a major part in preventing the taking of action against underrepresented forms of plastic pollution. In this paper, we conduct a computational analysis of nearly 1600 newspaper articles from 10 of the most popular media sources in the United States and India. Our analysis shows that despite increasing media coverage of plastic pollution, certain forms remain dangerously underreported. Additionally, there is a discernible difference between coverage in the two countries, suggesting the need for consumers to aggregate information to get a comprehensive view of the issue. Our use of a large language model – as opposed to a manual approach – allows us to conduct our analysis on a larger dataset in the future, while still retaining high classification accuracy.
Archival Submission: arxival
Submission Number: 32
Loading