Citation

Navarro J, Piña JU, Mas FM, Lahoz-Beltra R. Int. J. Disaster Risk Reduct. 2023; 91: e103694.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.ijdrr.2023.103694

PMID

unavailable

Abstract

In this work we have used as a source of information a large sample of the press articles published during 2021 about the eruption of the Cumbre Vieja volcano in the island of La Palma (Canary Islands). In contraposition, the scientific papers evaluating different facets of natural disasters have preferentially used social networks as a source of information. Herein we have shown how the emotions and sentiments expressed in press media can be efficiently analyzed via AI techniques to better assess the social impact of a disaster at the time it takes place. We have also gauged the usefulness of different classifiers combining sentiment analysis with multivariate statistical analysis and machine learning techniques. By applying this methodology, we were able to classify a newspaper article within a certain time frame of the eruption, and we observed significant differences between local news published in Spanish and those of foreign newspapers written in English. We also found different emotional trajectories of articles by applying the Fourier transform onto the inner "valence" progress along each article narrative time. In addition, there appeared a significant relationship between the surface area occupied by lava and the emotions and sentiments expressed in the articles--many other correlations and causalities could be explored too. The main findings of this research may constitute a helpful resource for a better understanding of the way press media react to volcanic activity, and may guide in public decision-making under different temporal horizons, including the design of improved strategies in the risk reduction domain.

Language: en

Keywords

Disaster journalism analysis; Machine learning; Sentiment analysis; Volcanic eruption management