Content area
The rise of artificial intelligence (AI) and "big data" has revolutionized the study of global conflicts, driving the development of tools to better understand their dynamics. While conflict datasets like GDELT and ACLED provide extensive event data, they often lack explicit causal information, limiting their effectiveness for nuanced analysis and prediction. This dissertation hypothesizes that enriching these datasets with structured causal insights, rather than focusing solely on algorithmic improvements, is a critical step toward advancing conflict analysis.
This research leverages innovative advances in Natural Language Processing (NLP), utilizing transformer-based models such as BERT and LLAMA2 to extract and categorize conflict causes from textual data. These models' ability to capture deep semantic relationships and contextual nuances makes them uniquely suited for identifying the multifaceted drivers of conflict.
To systematically classify conflict causes, the PMESII framework—encompassing political, military, economic, social, information, and infrastructure dimensions—is employed. By integrating PMESII with enriched textual data, this study captures complex causal patterns, enhancing the depth and utility of conflict datasets. Preliminary experiments demonstrate the feasibility and value of combining causal annotations with advanced NLP techniques, offering significant improvements in the organization and interpretability of conflict-related information.
The findings underscore the transformative potential of integrating structured frameworks like PMESII with state-of-the-art NLP models to enrich conflict datasets. This dissertation contributes a novel methodological approach for identifying and organizing conflict causes, setting the stage for more accurate, robust, and actionable conflict prediction models in the future.
