ZPT Knowledge Navigation Ontology

A semantic vocabulary for dynamic knowledge navigation using camera metaphors to control information access across three dimensions: Zoom (abstraction level), Pan (domain scope), and Tilt (representation method).

Status 2025-12-30 : aligned with Semem ZPT implementation; vocabulary refreshed and still evolving.

• Ontology Specification - RDF/OWL vocabulary in Turtle format
• Namespace Document - class structure & reference term descriptions
• ZPT for Librarians - human-oriented explanation with examples

There are also confusing diagrams under media/.

Agents, both human and artificial have limits on the amount of information they can process at a given time : working memory/context window. In attempt to optimise processing workflows, two complimentary ontologies have been developed: Ragno, which offers a representation of a knowledgebase that is friendly to both Semantic Web and Language Model technologies, and here ZPT which provides a means of purposefully navigating the knowledgebase.

Ragno describes a heterogenous knowledge graph in a generalized fashion, loosely scale-free - it may be a domain-specific local knowledgebase or the Web as a whole. ZPT adds navigation capabilities that adapt information presentation to user (agent) needs. A system may dynamically select optimal knowledge based on query requirements across three orthogonal dimensions:

• 🔍 Zoom: Controls abstraction level (micro → entity → text → unit → community → corpus)
• 🔄 Pan: Controls domain boundaries (topics, entities, temporal periods, geographic regions, corpuscle scopes)
• 📐 Tilt: Controls representation method (embeddings, keywords, graph structure, temporal, memory)

ZPT is being developed in parallel with Ragno and application tooling with Semem. Still early days.

• Dynamic Selection: Automatically chooses optimal information subsets for any query
• Multi-dimensional Optimization: Balances detail level, relevance, and representation effectiveness
• Provenance Tracking: Full navigation history using PROV-O integration
• Standards-Based: Built on SKOS, PROV-O, and OWL foundations
• Extensible: Domain-specific zoom levels, pan domains, and tilt projections

@prefix zpt: <http://purl.org/stuff/zpt/> .
@prefix ragno: <http://purl.org/stuff/ragno/> .

# Create a navigation view
ex:myView a zpt:NavigationView ;
    zpt:hasZoomState [ zpt:atZoomLevel zpt:UnitLevel ] ;
    zpt:hasPanState [ zpt:withPanDomain ex:aiTopic ] ;
    zpt:hasTiltState [ zpt:withTiltProjection zpt:EmbeddingProjection ] ;
    zpt:answersQuery "What are machine learning applications?" .

1. Query: "Explain neural networks for beginners"
2. Zoom: Community level (high-level summaries)
3. Pan: AI/ML topic domain
4. Tilt: Keyword projection (accessible language)
5. Result: Beginner-friendly overview corpuscles selected automatically

Query → ZPT Analysis → Corpuscle Selection → View Projection
  ↓         ↓              ↓                 ↓
Zoom     Pan           Optimization      Representation
Level   Domain         Algorithm         Method

ZPT is designed to work with:

• Ragno - Heterogeneous knowledge graph ontology
• NodeRAG - Graph-based retrieval-augmented generation
• SKOS - Concept organization and collections
• PROV-O - Provenance and navigation history

• Academic Research: Progressive topic exploration from overview to details
• Technical Documentation: Adaptive presentation based on user expertise
• News Analysis: Multi-scale information from headlines to source documents
• Legal Discovery: Consistent detail across case types for comparison

In use the navigation flow might look something like this, where User Query could come from any agent:

Current Version: 0.2.0 (Semem-aligned update)

• ✅ Core ontology design complete
• ✅ Integration with Ragno defined
• ✅ Standard vocabulary compliance (SKOS, PROV-O)
• 🔄 Reference implementation in progress
• 📋 Evaluation framework planned

See Development Plan for detailed specifications.

• NodeRAG Paper: Structuring Graph-based RAG with Heterogeneous Nodes
• GraphRAG: Microsoft's graph-based retrieval augmentation
• LightRAG: Optimized graph-based knowledge retrieval
• HippoRAG: Neurobiologically-inspired knowledge graphs

Contributions welcome! Areas of interest:

• Domain-specific zoom level definitions
• Novel tilt projection methods
• Optimization algorithm improvements
• Integration with existing knowledge systems
• Evaluation metrics and benchmarks

• Code: MIT License
• Documentation: CC BY 4.0 (Creative Commons Attribution)
• Ontology: CC BY 4.0 (Creative Commons Attribution)

Creator: Danny Ayers
Homepage: https://danny.ayers.name
Repository: https://github.com/danja/zpt

@misc{ayers2025zpt,
  title={ZPT: Zoom-Pan-Tilt Knowledge Navigation Ontology},
  author={Ayers, Danny},
  year={2025},
  url={http://purl.org/stuff/zpt/},
  note={Version 0.2.0}
}

ZPT enables intelligent knowledge navigation through dynamic adaptation of information presentation to user needs and query requirements.