Project: RAG-based Research Assistant

[Jwizzed]

Project Name

PaperMaid

Description

RAG-based Research Assistant

Description

This Retrieval-Augmented Generation (RAG) system is designed to help researchers and academics efficiently summarize and understand multiple research papers or conference proceedings. By leveraging advanced AI and database technologies, our application streamlines the process of analyzing and synthesizing information from various scientific sources.

Our system maintains a database of research papers in Azure Cosmos DB. When a user uploads a file, the application performs PDF extraction on the document and conducts a web search for supplementary information. This extracted content and web search results are combined with the user's prompt. The combined information is then embedded and used to find relevant papers in the Cosmos DB through vector similarity search. These relevant papers from the database are used to add more context to the user's query. All this information is sent to a Large Language Model (LLM) to generate a comprehensive response, which includes a knowledge graph for easier visualization of the relationships between papers and concepts.

Key Features:

• Multi-document summarization
• Knowledge graph generation for easy comprehension
• Identification of relationships between papers
• Web search integration for additional context
• PDF extraction for user-uploaded documents

Data Sources:

• Pre-existing research papers in Cosmos DB
• User-uploaded research papers and conference proceedings
• Web search results for supplementary information

Technical Stack:

• Backend: Langchain with Azure OpenAI
• Database: Azure Cosmos DB for NoSQL
• Vector Search: Azure Cosmos DB Vector Search
• Embeddings: Langchain Embeddings
• Frontend: Gradio
• Caching: Azure Cosmos DB (for chat history)

Modifications:

• Custom prompt engineering for research paper analysis
• Integration of Azure Cosmos DB Vector Search for efficient similarity matching
• Implementation of knowledge graph generation for visualizing relationships between papers
• PDF extraction and web search integration for comprehensive context building

Target Audience:

• Researchers and academics across various disciplines
• Graduate students conducting literature reviews
• Research institutions and universities
• Anyone needing to quickly grasp and synthesize information from multiple scientific papers

This application aims to significantly reduce the time and effort required to process and understand large volumes of scientific literature, enabling users to focus on drawing insights and advancing their research by leveraging both existing knowledge and new inputs.

Technology & Languages

• [ ] JavaScript
• [ ] Java
• [ ] .NET
• [X] Python
• [ ] AI Studio
• [ ] AI Search
• [ ] PostgreSQL
• [X] Cosmos DB
• [ ] Azure SQL

Project Repository URL

https://github.com/papermaid/papermaid

Deployed Endpoint URL

No response

Project Video

https://www.youtube.com/watch?v=QGih0oOmHHs

Team Members

Krittin Setdhavanich, Sirin Puenggun, Chaiyawut Thengket

[Yash-Kavaiya]

@Jwizzed Explanation is demo is Excellents

[multispark]

Hello @jwizzed, thank you for participating in RAG Hack!

The team is working hard to distribute badges. Please have each team member fill out this form: aka.ms/raghack/badge-dist

Thank you!