Replication Data and Code for: Identifying Macro-Financial Signals in Bank Indonesia's Communication Using Transformer-Based Semantic Clustering

This repository contains the textual corpus, processed regression dataset, and Python replication code for the paper 'Identifying Macro-Financial Signals in Bank Indonesia's Communication Using Transformer-Based Semantic Clustering.' The corpus comprises 230 monthly monetary policy executive summaries from Bank Indonesia (November 2005 – March 2026). The replication code covers text preprocessing, CentralBankRoBERTa embedding extraction, boilerplate penalization, K-Means clustering, cluster validation, and Local Projections estimation with Newey-West HAC standard errors.