End-to-End Retrieval-Augmented Generation (RAG) System

👉 Link to the RAG System

This is the project repository for the course DSAN5800: Advanced NLP at Georgetown University, 2024 Fall.

Group Members

• Yuting Fan (yf265@georgetown.edu)
• Peng Li (pl800@georgetown.edu)
• Yiwei Qi (yq125@georgetown.edu)

Description:

In this project, you will build a retrieval-augmented generation (RAG) system that can retrieve relevant documents from a knowledge base and use a language model to generate responses based on the retrieved information.

Key Tasks:

• Implement a document retriever using dense retrieval (e.g., FAISS).
• Combine the retriever with a language model (e.g., GPT) to generate responses.
• Build a simple web application that satisfies these functional requirements:
  • Chat interface
  • Document upload
  • Document source attribution
• Evaluate retrieval and response generation quality.

Resources:

• HuggingFace RAG Implementation
• Google Drive Workspace
• LangChain: Build a Retrieval Augmented Generation (RAG) App
• RAG Evaluation

Project Deliverables:

• A well-documented GitHub repository containing your code
• A project report (4-6 pages) outlining your methodology, experiments, and results
• A 15 minute final presentation / demo showcasing your project, followed by a 5 minute Q&A session