EINDICE / Financial-Sentiment-Analysis-with-ChatGPT

This repository contains the code and supplementary materials for our research paper on financial sentiment analysis, where we explore the capabilities of ChatGPT 3.5 in the context of the foreign exchange (forex) market.

Financial Sentiment Analysis with ChatGPT

This repository contains the code and supplementary materials for our research paper on financial sentiment analysis, where we explore the capabilities of ChatGPT 3.5 in the context of the foreign exchange (forex) market.

• Paper Title:Transforming Sentiment Analysis in the Financial Domain with ChatGPT
• Authors: Georgios Fatouros, John Soldatos, Kalliopi Kouroumali, Georgios Makridis and Dimosthenis Kyriazis
• Submitted to: Elsevier Machine Learning with Applications
• Cite as: G. Fatouros, J. Soldatos, K. Kouroumali et al., Transforming sentiment analysis in the financial domain with ChatGPT. Machine Learning with Applications (2023), doi: https://doi.org/10.1016/j.mlwa.2023.100508.

Prerequisites

• Python 3.x

Setup

1. Clone the repository to your local machine.
2. Install the required Python libraries using pip:

   pip install requirements.txt

Data

1. sentiment_annotated_with_texts.csv: Sentiment Annotated dataset of Forex news. Includes sentiment predictions from FinBERT. For more info check https://zenodo.org/records/7976208.
2. prompts.json: Includes the different prompts presented in the paper.
3. sentiment_predictions_single_article.csv: Contains the predictions from the prompts processing single headlines.
4. sentiment_predictions_allday_articles.csv: Contains the predictions from the prompts processing all the headlines per day (GPT-P5 and GPT-P6).

Running the prompts.py

To run the script and evaluate ChatGPT's performance on a sample row from the dataset.

Add your OpenAI API key to the prompts.py script: Replace in the script with your actual OpenAI API key.

python3 prompts.py

The script will print the sentiment predictions of ChatGPT for the specified row and compare them against the true sentiment and FinBERT's predictions.

Running the results.py

The script will print the comparative performance results of the models/prompts presented in the paper .

python3 results.py

Helper Functions

sentiment_to_numeric: Converts sentiment labels to numeric values. get_sentiment: Fetches sentiment predictions from ChatGPT based on the specified prompt.

License

This project is open-source and available under the MIT License.

Acknowledgment

Part of the research leading to the results presented in this paper has received funding from the European Union’s funded Project FAME under grant agreement no 101092639.

We would also like to express our gratitude to our FAME partners, JRC Capital Management Consultancy Research GmbH and KMcube Asset Management SA, for their invaluable contributions and expertise in the data labeling process.