Analyzing Israeli Cabinet Deliberations: A Computational Workflow and Preliminary Results

How do wars transform the Israeli government's agenda? Historians have long answered such questions with traditional qualitative tools. Computational methods, including LLMs and machine learning, now enable research on a broader scale with quantitative findings.

This paper presents the first stage of a computational corpus of all Israeli cabinet deliberations (1948–1995), using the Thirteenth Government (1966–1969), which served before and after the Six-Day War, as a pilot. Our workflow converts the State Archives' scanned minutes into structured data through OCR, segmentation, metadata, and BERTopic modeling with multilingual-e5-large embeddings. LLM-assisted refinement and expert review then expanded the initial model output into 24 topics under seven parent categories, validated through fixed-category coding.

Our contributions are twofold: methodologically, a reproducible workflow and a reusable database and dashboard for other historians; substantively, an empirical account of wartime agenda change. For example, economic issues fell from 50% of discourse before the war to 18% afterward.