HyperKnowledge/ MMM components

Work within a domain application

• Local concept data, connected to domain data.
• Local concepts use any model, expose local Ids.
• Domain-specific view of domain concepts

Connect domain applications through hub

• Abstract KHub model for many concept models
• Unification can use NLP or Humans. Contestable.
• Domain application links to community graph view

Concept Identification

… in external domain data & concept descriptions.

• Contestable suggestions by machine or humans
  • Later: Reputation layer

Transclusion

• Embed a domain view in the KHub interface
• Embed KHub views in the Domain applications
  • Text view with embedded concepts

Concept Unification as a Service

KHub federation

• Federated queries on KHubs
• Negotiate global identifiers (→ Canonicalization)

Socio-Epistemic federation

• Epistemic status simple in community
  • Eg: research question vs provisional truth
• Federated status: Is statement contested? etc.

Unification Reputation mechanisms

Reputation for sensible distinctions (≠truth!)

Subscription mechanisms

• KHub concept sub → Live views in Domain Apps
• Federation sub → Live KHubs (wrt center)
• Domain app sub → Live KHubs (wrt periphery)

Short term

1. Domain-specific concept UX
2. Abstract concept model (Topic Maps)
3. Contestable Concept unification model
4. Build the KHub accordingly
5. View of community concepts
6. Basic unification/distinction UX
7. Links from domain application

Mid-term

1. Identify concepts within domain objects
2. Local epistemic status
3. Transclusion in domain app
4. Improve local map for community embedding
5. Improve unification UX
6. Experiments with NLP unification
7. Unification validation by humans

Long term

1. Federated queries between KHubs
2. Global Concept Identifiers by contagion
3. Canonicalization
4. Epistemic federation
5. Unification reputation
6. Epistemic reputation