Year: 2025
Type: conference-paper
Abstract: We used data from LexisNexis to determine the words in news media that were most important in classifying countries as higher or lower peace using logistic regression, support vector machines, decision trees, and random forest as classifiers. We then embedded those words using the multilingual-e5-large-instruct model, obtaining high-dimensional representations of each word’s semantic meaning. PCA allowed us to transform the word embeddings into a two-dimensional space, facilitating visual interpretation of the relationships and clusters among the words. With the data transformed into a 2D space, we applied K-means clustering to identify distinct groups or clusters of words that shared similar characteristics. We found that higher peace news is characterized by themes of finance, daily activities, and health and that lower peace news is characterized by themes of politics, government, and legal issues. This work provides a starting point to measure levels of peace and identify the social processes that underly those words.
Authors: Tushar Prasad, Larry S. Liebovitch, Melissa Wild, Harry West, Peter T. Coleman
Institution: Columbia University
Language: English
Subfield: General Social Sciences
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
Funder: Columbia University
OpenAlex ID: W4409156316