Words that Represent Peace

TL;DR

The paper investigates how media language correlates with societal peace by analyzing English-language articles across 20 countries using LexisNexis data. It labels countries as peaceful or non-peaceful by aggregating multiple peace indices and trains four supervised classifiers with a leave-one-country-out scheme to predict peace status from word features. It combines quantitative modeling with qualitative analyses—word clouds, word embeddings reduced by PCA, clustering, and LLM-based semantic segmentation—to reveal thematic differences: peaceful discourse centers on finance, daily life, and health, while non-peaceful discourse centers on politics and conflict. The work demonstrates a data-driven method to quantify peace through language and suggests potential applications for journalism and peace-building research.

Abstract

We used data from LexisNexis to determine the words in news media that best classifies countries as higher or lower peace. We found that higher peace news is characterized by themes of finance, daily actitivities, 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.

Figures (6)

• Figure 1: Random Forest Peaceful Word Cloud
• Figure 2: Random Forest Nonpeaceful Word Cloud
• Figure 3: SVM Peaceful Word Cloud
• Figure 4: SVM Nonpeaceful Word Cloud
• Figure 5: SVM K-means Clustering