Building and Retrieval of Knowledge Graph from Documentation and graph rag
Debrup Paul (~debrup) |
Description:
Project Overview: This project focuses on developing a hybrid summarization system that leverages both Abstractive and Extractive approaches, built on foundation models. The primary goal is to create a production-grade system capable of summarizing GitHub repositories. It spans the entire development cycle, from data collection and model experimentation to final deployment. A key challenge is achieving high performance with smaller models, contributing to SCoRe Lab’s portfolio of AI-driven solutions.
New Feature: GraphRag Module This year, we introduced the GraphRag module, designed to build and retrieve knowledge graphs from project documentation. Key benefits include:
- Ease of Use for Beginners: Simplifies the process for newcomers exploring new libraries.
- Seamless Library Switching: Provides flexibility to switch between libraries without needing to review entire documentation, such as with pandas.
- Timely Adaptation: Addresses the gap in knowledge for new and smaller projects that tools like ChatGPT may not yet cover.
- Debugging Support: Enhances debugging processes by making information retrieval from documentation more efficient.
Prerequisites:
A basic understanding of Graphs, large language models(LLMs), and familiarity with libraries such as HuggingFace, Ollama, Ragas, and LlamaIndex is recommended. Participants should also have a working knowledge of Python, particularly object-oriented programming (OOP).
While these concepts are beneficial, I will aim to make the session as beginner-friendly and intuitive as possible.
Speaker Info:
Debrup Paul is a Google Summer of Code contributor at C2SI and is currently pursuing an M.Sc. in Mathematics along with a B.E. in Mechanical Engineering at BITS Pilani, Goa Campus. With a strong interest in Generative AI and Natural Language Processing (NLP),he enjoys reading research papers in the domains of large language models (LLMs) and NLP and building products for community. Debrup is passionate about exploring the frontiers of AI technology and contributing to its advancement.