Evaluating The Effectiveness of The Llama2 Large Language Model in Analyzing Economic Texts: A Comparative Analysis Using Diverse Data Sources

Abstract

This study investigates the effectiveness of the open-source llama2 large language model in analyzing various types of economic texts. We employ a comparative analysis approach, utilizing data from four diverse sources: Federal Reserve Economic Data (FRED), Edgar (EDGAR) Database from the U.S. Securities and Exchange Commission (SEC), International Monetary Fund (IMF) Data, and World Bank Open Data. We focus on the performance of llama2 for specific tasks like sentiment analysis, entity recognition, and topic modeling. The findings will contribute to understanding the potential and limitations of using large language models for extracting insights from diverse economic data sources.