Masters Thesis Research Project: A Novel Approach to Improve Question Answering System using GPT-3 🎓🔍

Overview 📜

This repository showcases my Masters Thesis Research Project where I delved into innovative techniques to enhance Question Answering Systems. The cutting-edge models BERT and OpenAI GPT-3 were employed, and their performance was evaluated using the Stanford Question Answering Dataset (SQuAD).

Objectives 🎯

• Investigate the efficacy of BERT and GPT-3 in the domain of Question Answering.
• Analyze accuracy and loss metrics for both models.
• Compare the performance of BERT and GPT-3 based on SQuAD evaluations.

Key Features 🛠️

• Data Preprocessing: Adapted the SQuAD dataset to suit the model requirements.
• Model Training: Trained BERT and GPT-3 using the processed SQuAD dataset.
• Evaluation Metrics: Calculated accuracy and loss for performance assessment.
• Comparative Analysis: Comprehensive comparison of BERT and GPT-3 performance metrics.

Installation 🖥️

1. Clone the repository:

git clone https://github.com/rhakbari/ms-thesis-gpt-3.git

2. Navigate to the project directory:

cd ms-thesis-gpt-3

Usage 📊

1. Data Preparation: Ensure SQuAD dataset is correctly located.
2. Model Training: Execute training scripts for BERT and GPT-3.
3. Evaluation: Run evaluation scripts for accuracy and loss calculations.
4. Comparison: Use analysis tools for BERT vs. GPT-3 performance assessment.

Results 📈

Detailed insights into BERT and GPT-3 performance metrics, including graphical representations and statistical analyses.

Conclusion 📝

Summarization of research findings, highlighting strengths and limitations of BERT and GPT-3 in Question Answering Systems.

Future Work 🔮

Discussion on potential research directions and methodology enhancements.

Contributors 👥

• Raza Hussain Akbari

For inquiries or feedback, please reach out.