Generative-Chatbot

This repository contains a Jupyter notebook designed to train a generative chatbot using the Ubuntu Dialogue Corpus. The dataset, available at: https://www.kaggle.com/datasets/rtatman/ubuntu-dialogue-corpus, consists of conversations extracted from Ubuntu chat logs, typically used for Ubuntu technical support.

Notebook Overview

Key Features:

1. Model Development:

• Implementation of two sequence-to-sequence models:
  • One with an attention mechanism (Luong or Bahdanau).
  • One without attention.
• Use of PyTorch for model training and optimization.

2. Data Preprocessing:

• Cleaning and tokenizing conversation text.
• Create dialogue pairs to mimic conversations.
• Splitting the data into training and testing sets.

3. Evaluation Metrics:

• BLEU and METEOR scores for measuring the quality of generated responses.
• Cosine similarity for response evaluation.

4. Libraries Used:

• PyTorch
• TensorFlow (Tokenizer)
• NLTK (BLEU and METEOR scores)
• Scikit-learn
• Pandas
• NumPy

Requirements

To run the notebook, ensure the following Python libraries are installed:

• PyTorch
• TensorFlow
• NLTK
• Scikit-learn
• Pandas
• NumPy
• tqdm

Install the required libraries using:

pip install torch tensorflow nltk scikit-learn pandas numpy tqdm

How to Use

1. Clone the repository:

git clone https://github.com/Vimal-Raghubir/Generative-Chatbot.git

2. Open the Jupyter notebook:

jupyter notebook Generative_Chatbot.ipynb

3. Follow the structured steps in the notebook to:

• Preprocess the dataset.
• Train and evaluate the models.
• Generate chatbot responses.

Performance Evaluation

The notebook evaluates model performance using:

• BLEU scores for n-gram overlap between generated and reference responses.
• METEOR scores for semantic similarity.
• Cosine similarity for vector-based evaluation.

Future Enhancements

Potential improvements include:

• Expanding training with larger models or additional datasets.
• Implementing Transformer-based architectures like GPT or BERT for enhanced response quality.
• Fine-tuning the attention mechanism for better context understanding.

Acknowledgments

This project is inspired by the Ubuntu Dialogue Corpus and aims to demonstrate the development of generative chatbots for technical support. Special thanks to Kaggle for providing the dataset.