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.
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
-
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