ABSTRACT
Social media has become an important means of communication and new insights can be gained from processing this data on a large scale. Our goal is to develop and implement a pipeline to automatically extract and analyse Twitter data on natural disasters and environmental topics. We aim to provide an additional layer of spatiotemporal data that can be used to study the immediate and lasting impacts of natural disasters, climate change, and environmental topics on the global population. An initial analysis of forest fires was conducted in four different languages confirming the need for multilingual support for global analysis. We found a positive correlation between wildfire occurrence and tweeting behaviour, as well as the geographic spread of fires. We found that simple sentiment predictions add little value when aggregating data on a large scale. A subsequent test using a fine-tuned stance detection model proved promising in determining the stance of tweets towards nuclear energy. We intend to expand our dataset and develop customised models in the future that can be used to analyse the global impact of natural disasters and environmental topics.
Acknowledgments
The research activities as described in this paper were funded by Ghent University and The Research Foundation - Flanders (FWO) (Grant number: G0F2820N).
Disclosure statement
No potential conflict of interest was reported by the author(s).