Abstract:
|
How can we tell when the U.S. grand strategy has shifted? Typically, analysis of a state's grand strategy - materialized through leaders' speeches, policies, and other government records - is done manually by reading and/or hand-coding these documents. To more efficiently and accurately evaluate these changes, we propose a new, automatic and unsupervised approach using topic modeling and statistical natural language processing.
Using a dataset containing over 6,000 speeches given by Jimmy Carter, Ronald Reagan, and George H.W. Bush over the course of their presidencies, we apply topic modeling and sentiment analysis in an attempt to illuminate the structure of Presidential discourse and identify markers of change in grand strategy. We employ a number of topic modeling tools including structural topic modeling (STM) and topics over time (TOT) to characterize the evolution of topics within each Presidency and identify factors that may predict change in strategy. Diving deeper, we apply similar methods to a subset of the data that is directly related to global terrorism in order to assess how conversation changes in reaction to terror incidents.
|