Nowcasting economic activity through text mining: evidence from Armenia
Central Bank of Armenia
Nowcasting economic activity through text mining: evidence from Armenia
Discussion paper DP-2026-01
Manishak Grigoryan
Abstract
This paper examines the application of text-mining techniques to economic news data for improving short-term macroeconomic assessment and nowcasting in Armenia. Drawing on international central bank practices, including those of the Bank of England, the Federal Reserve, and the Central Bank of Nigeria, the study highlights the growing role of machine learning and natural language processing (NLP) in extracting policy-relevant signals from unstructured textual data. Using a dataset of over 100,000 economic news articles, the paper applies a combination of word frequency analysis, co-occurrence and clustering methods, and sentiment and uncertainty indicators to identify key economic themes and dynamics. The results show that news-based indicators capture meaningful information about economic activity, particularly in relation to trade, geopolitical developments, and sectoral trends. The findings suggest that text-based measures can complement traditional macroeconomic models and provide timely insights for monetary policy analysis in data-constrained environments.