Replication package for narrative forecast extraction and benchmark evaluation from bank research reports
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This archive contains anonymous replication materials for a manuscript that studies how structured forecasts can be extracted from narrative bank research reports and evaluated against benchmark time-series models. It includes the raw text corpus, raw numeric macroeconomic and financial series, the archived structured forecast dataset used in the main replication, manual-review and gold-standard validation inputs, and Python code with uv environment files needed to reproduce the reported tables and figures. The main replication can be run without any API access by starting from the archived structured forecast file included in the package. The full raw-text extraction pipeline is also included in code and can be rerun separately with an external API key. Main inputs are Results/reports_raw_final.csv, Data/PL_monthly.csv, Data/eurpln_d.csv, Results/structured_forecasts_final.csv, Results/fdf_large_errors_verified.csv, and Data/Golden_standard_coding_done.xlsx. Main commands are uv sync and ./run_replication.sh. Optional raw-text extraction can be run with ./run_optional_raw_text_extraction.sh after setting OPENAI_API_KEY. The archive intentionally excludes manuscript files and project-author identifiers.
Fields of Study & Subjects
Forecasting Techniques and Applications
Computational and Text Analysis Methods
Stock Market Forecasting Methods
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 28 March 2026
- Language
- EN
- Document type
- Dataset