Skip to content

Zpankz/Knowledge-Graph-Chat

Repository files navigation

Literary Knowledge Graph RAG

A sophisticated system that builds narrative-aware knowledge graphs from literary texts and enables natural language querying using Graph RAG (Retrieval Augmented Generation) with Chain of Thought reasoning.

Theory & Background

This system implements several key concepts:

  • Graph-based RAG: Extends traditional RAG by structuring information in a knowledge graph, preserving relationships and context
  • Chain of Thought (CoT) Reasoning: Uses guided walks through the graph to build coherent narrative understanding
  • Community Detection: Identifies thematically related subgraphs to improve retrieval relevance
  • Narrative Coherence: Maintains story flow and context through specialized graph traversal

Core Features

  • Intelligent Graph Construction

    • Automatic entity and relationship extraction
    • Narrative-aware community detection
    • Dynamic graph improvement through CoT analysis
  • Advanced Query Processing

    • Graph-guided retrieval
    • Multi-hop reasoning
    • Context-preserving response generation
    • Narrative coherence maintenance
  • Interactive Features

    • Natural language chat interface
    • Graph visualization
    • Detailed analysis traces
    • Source attribution

About

No description or website provided.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages