Eliminating ontology contradictions based on the Myerson value

Area of Science:

• Artificial Intelligence
• Knowledge Representation
• Ontology Engineering

Background:

• Ontologies are crucial for AI knowledge representation but often contain logical contradictions.
• Traditional Integer Linear Programming (ILP) models treat all formulas equally, failing to prioritize critical ones for contradiction resolution.

Purpose of the Study:

• To develop an advanced ILP model for ontology contradiction resolution that prioritizes formula deletions.
• To integrate cooperative game theory and graph-based representations for more effective contradiction management.

Main Methods:

• Computed Shapley values to quantify the marginal contribution of each formula in resolving contradictions.
• Extended Shapley values to Myerson values using a graph-based ontology representation.
• Developed a Myerson-weighted ILP model for lexicographic elimination of logical contradictions, minimizing deletions and prioritizing based on Myerson values.

Main Results:

• The proposed Myerson-weighted ILP model preserves more graph edges compared to traditional ILP methods.
• The approach effectively quantifies formula contributions and establishes clear deletion priorities.
• Demonstrated superior performance across 18 diverse ontologies.

Conclusions:

• The novel approach offers a more nuanced and effective method for resolving logical contradictions in ontologies.
• This technique enhances ontology accuracy and data integrity by prioritizing deletions based on calculated contributions.