RAG-based Restaurant Chatbot

A Retrieval Augmented Generation (RAG) based chatbot that answers user questions about restaurants using data scraped from restaurant websites.

Project Overview

This project consists of the following components:

1. Web Scraper: Collects restaurant data including menus, locations, operating hours, etc.
2. Knowledge Base: Processes and stores scraped data for efficient retrieval
3. RAG Chatbot: Uses Hugging Face models to retrieve and generate responses
4. User Interface: Streamlit app for interacting with the chatbot

Setup Instructions

Prerequisites

• Python 3.8+
• Git

Installation

1. Clone the repository:

   git clone https://github.com/keyaaness/zomato-rag-chatbot.git
   cd zomato-rag-chatbot
   

2. Create a virtual environment (optional but recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install the required packages:

   pip install -r requirements.txt
   

Usage

Running the Web Scraper

To collect restaurant data:

python src/scraper/main.py

This will scrape data from the configured restaurant websites and save it to data/raw/.

Building the Knowledge Base

To process the scraped data and build the knowledge base:

python src/knowledge_base/build_kb.py

This will process the raw data and create a structured knowledge base in data/processed/.

Running the Chatbot Interface

To launch the Streamlit interface:

streamlit run src/app.py

Then open your browser and navigate to http://localhost:8501 to interact with the chatbot.

Project Structure

zomato-rag-chatbot/
├── data/                      # Data directory
│   ├── raw/                   # Raw scraped data
│   └── processed/             # Processed data for knowledge base
├── src/                       # Source code
│   ├── scraper/               # Web scraping module
│   ├── knowledge_base/        # Knowledge base processing
│   ├── rag/                   # RAG implementation
│   ├── utils/                 # Utility functions
│   └── app.py                 # Streamlit application
├── notebooks/                 # Jupyter notebooks for exploration
├── tests/                     # Test files
├── README.md                  # Project documentation
└── requirements.txt           # Package dependencies

Features

• Scrapes data from multiple restaurant websites
• Structured knowledge base for efficient retrieval
• Natural language query processing
• Conversational interface via Streamlit
• Handles various types of restaurant-related queries
• Remembers conversation context

Limitations

• Only works with data from scraped restaurants
• May not handle highly complex or ambiguous queries
• Limited to text-based information (no image processing)

Future Improvements

• Expand the number of restaurants in the database
• Implement more advanced retrieval mechanisms
• Add sentiment analysis for restaurant reviews
• Improve handling of ambiguous queries
• Create a mobile application interface