Contextual Representation And Reasoning With Description Logics.pdf

- Objective: To qualify knowledge with context in the Semantic Web (SW).
- Approach: Adopts McCarthy's "context as a box" metaphor, representing context as a logical statement "box" bounded by contextual attributes (dimensions).
- Key Requirements:
- Knowledge about context: Explicitly represent and reason about contextual dimensions and relations.
- Contextually bounded facts: State local effects that don't propagate universally.
- Reuse/lifting of facts: Automatically include information from specific contexts into more general ones without spoiling locality.
- Overlapping and varying domains: Objects can be in multiple contexts but not all.
- Inconsistency tolerance: Allow for potentially conflicting facts within contexts without system-wide inconsistency.
- Complexity invariance: Qualification of knowledge by context should not increase computational complexity.

Framework:

- Syntax & Semantics: Defined based on the expressive description logic SROIQ, allowing for hierarchical coverage relations between contexts.
- Computational Properties:
- Concept satisfiability and subsumption are decidable with a 2NEXPTIME upper bound (same as SROIQ).
- Sound and complete Natural Deduction calculus for knowledge propagation between contexts.

Meta-Language & Object-Language:

- Meta-language: Uses Γ, a vocabulary including context identifiers, dimensions, dimensional values, and coverage relations. Specifies context structure and properties.
- Object-language: Used within each context to specify knowledge, distinct from the meta-language to maintain good computational properties.