https://github.com/mahbub2001/physics-chatbot

https://github.com/mahbub2001/physics-chatbot

Last synced: 2 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/mahbub2001/physics-chatbot
• Owner: Mahbub2001
• Created: 2024-12-09T19:14:05.000Z (about 2 months ago)
• Default Branch: main
• Last Pushed: 2024-12-09T19:23:51.000Z (about 2 months ago)
• Last Synced: 2024-12-09T20:28:17.474Z (about 2 months ago)
• Language: Python
• Size: 4.91 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Physics Book Chatbot

## **Physics Book Chatbot**

A physics-focused chatbot application built using **Streamlit** and **LangChain**, designed to answer questions strictly based on a provided physics textbook PDF. The chatbot also includes an evaluation pipeline to measure the quality of its answers using **ROUGE** and **BLEU** metrics. It uses the **Llama 3.1** model for generating accurate, context-based responses.

---

## **Features**

1. **Physics Chatbot:**

   - Answers questions based solely on the content of the physics textbook.

   - Retrieves relevant information using embeddings and a retriever.

   - Ensures no hallucinated or external information is included in responses.

2. **Llama 3.1 Model:**

   - Utilizes **Llama 3.1**, a state-of-the-art lightweight LLM optimized for high-quality, context-aware responses.

3. **Evaluation Metrics:**

   - Supports **ROUGE** (ROUGE-1, ROUGE-2, ROUGE-L) for overlap-based evaluation.

   - Implements **BLEU** for assessing response accuracy against reference answers.

   - Automatically processes a set of questions and reference answers for evaluation.

4. **Streamlit Interface:**

   - Simple and interactive chat interface.

   - Displays dynamic chatbot responses with a typing effect.

   - Presents evaluation results in an organized table format.

---

## **Evaluation Metrics**

### **ROUGE**

Measures the overlap between chatbot responses and reference answers:

- **ROUGE-1 F1**: Unigram overlap.

- **ROUGE-2 F1**: Bigram overlap.

- **ROUGE-L F1**: Longest common subsequence.

### **BLEU**

Evaluates the n-gram precision of chatbot responses against reference answers:

- Scores range from **0** (poor match) to **1** (perfect match).

---

## **Dependencies**

All dependencies are listed in `requirements.txt`. Key libraries include:

- **`streamlit`**: For the interactive web interface.

- **`langchain`**: For building the chatbot's retrieval-augmented generation (RAG) pipeline.

- **`chroma-db`**: For vector storage and retrieval.

- **`nltk`** and **`rouge-score`**: For evaluation metrics.

- **`Llama 3.1`**: For generating context-aware responses.

---

## **Acknowledgments**

- **LangChain**: For the framework enabling RAG-based conversational AI.

- **Ollama**: For the lightweight and efficient LLM used for generating responses.

- **Streamlit**: For providing an easy-to-use interface for interactive applications.

---

## **Setup and Installation**

### Prerequisites

- Python 3.8 or higher

- A physics textbook PDF (`phybook10.pdf`) 

- [Ollama LLM](https://ollama.ai/) running locally

- **Llama 3.1** installed and ready to serve responses

### Installation Steps

# Install additional packages using pip

pip install -r requirements.txt

# Run

streamlit run emne.py

# Test

streamlit run test_emne.py