Ragno is an ontology for describing heterogenous knowledge bases and their graph structures, designed for systematic, efficient and flexible retrieval and augmentation.

Status: 2025-05-28 : version 0.3.0. It should be adequate to begin testing in experimental deployment. Docs currently mostly AI-generated, manual intervention required.

I had a rough draft of a model in mind, which I'd been just-in-time developing for Semem "Semantic Memory", but then stumbled on the paper NodeRAG: Structuring Graph-based RAG with Heterogenous Nodes which was a remarkably close fit, so I've integrated some ideas from there.

• Term Reference
• Definition - Turtle RDF

Ragno extends SKOS (Simple Knowledge Organization System) to model heterogeneous knowledge graphs with various element types that support semantic retrieval and graph-based navigation.

Namespace : <http://purl.org/stuff/ragno/>

• Element - Base class for all knowledge graph components (subclass of skos:Concept)
• Corpus - A body of knowledge represented as a skos:Collection of Elements
• Corpuscle - A conceptual subset of a Corpus, also a skos:Collection

• Entity - Named entities serving as knowledge anchors
• Relationship - First-class connections between entities
• Unit - Semantic units representing independent concepts
• Attribute - Properties of important entities
• TextElement - Original text chunks with explicit content
• CommunityElement - Cluster within graph
• IndexElement - Search/retrieval-oriented structures (embeddings, keywords)

• Support for standard RDF/OWL tooling, graph algorithms, similarity techniques, LLM-friendly
• Scale-independent knowledge organization
• SKOS integration for concept organization
• PROV-O integration for provenance tracking