The Impact of Prompt Engineering on Large Language Models in Argument Mining

---

Overview

This repository contains code, data, and documentation for research on the impact of prompt engineering on large language models in the field of argument mining.

---

Table of Contents

• Repository Structure
• Installation & Setup
• Using the LLM
• Contact

---

Repository Structure

• batch_api/ – Data for the OpenAI Batch API
  • input/ – Input data
  • output/ – Output data
• data/ – Datasets
  • original/ – Argument-Annotated-Essays dataset (Version 2) by Stab and Gurevych (2017a), downloaded here
  • transformed/ – Processed data
• graphs/ – Visualizations and related notebooks
• prompts/ – Used prompts
  • building-blocks/ – Prompt building blocks
  • final-prompts/ – Final prompt variants
• report/ – Data-based report
• src/ – Python modules and helper functions
• 1.EDA.ipynb – Exploratory data analysis
• 2.data-transformation.ipynb – Data processing
• 3.llm.ipynb – Prompt creation & LLM queries
• 4.evaluation.ipynb – Evaluation of LLM results
• requirements.txt – Dependencies

---

Installation & Setup

It is recommended to use a virtual environment (e.g., Anaconda). Install the dependencies as follows:

conda create -n argument-mining python=3.12 -y
conda activate argument-mining
pip install -r requirements.txt

---

Using the LLM

To use the GPT-4o mini model via the OpenAI Batch API, an OpenAI API key is required. Store this in a .env file as OPENAI_API_KEY. More information about the API and key can be found here.

---

Contact

For questions or feedback:

• Email: datadrvn.ai@gmail.com

---