Unraveling the Complexities of Climate Change and Environment Migration: A Transformers-Based Topic Modelling Approach

Abstract

Abstract This research aims to synthesize the theoretical field on climate migrations by identifying the main thematic lines that make up the area of study, as well as to analyze their temporal evolution, geographic distribution, and impact. For this purpose, a thematic analysis of the abstracts of 1,048 scientific articles has been carried out by applying natural language processing techniques. The analyses consisted of the application of a clustering strategy based on high dimensionality vectors generated from the texts through the application of neural networks based on BERT architecture. The results show that the research on climate migrations is composed of a total of 15 distinct themes. It has also been found that each thematic line is different in their volume, temporal evolution, geographic distribution, and impact. This knowledge offers a privileged position to strengthen the development of the discipline by providing greater perspective to researchers and knowledge about the field of study itself.