The evolution of frequency, language complexity and policy narrative: benchmarking Python-based NLP and ChatGPT on a century of SNB rhetoric

We compiled and processed a novel, comprehensive dataset of 1,063 speeches delivered by high-ranking Swiss National Bank officials between 1928 and 2024. Beyond tracking their quantitative evolution, we conducted an in-depth analysis of the length, readability and thematic content. To ensure robust results, we benchmarked traditional Python-based NLP text mining against generative artificial intelligence using ChatGPT. Our findings highlight four primary policy insights. First, the frequency of annual speeches has followed an upward trajectory since the mid-1950s, accelerating notably after the transition to flexible exchange rates. Second, a clear shift toward brevity is evident. Third, the language has become increasingly accessible since the 1960s. Fourth, inflation has moved to the forefront of the thematic discourse in the era of flexible exchange rates, while cyclical economic development and unemployment have received less emphasis. Finally, we outline key methodological lessons for deploying LLMs to analyze central bank communication, emphasizing the complementary relationship between probabilistic language models and deterministic Python frameworks, while addressing inherent algorithmic pitfalls and identifying critical open questions for future research.