New Framework for Handling Contradictions in Knowledge Bases Using Weighted Costs

When knowledge bases contain contradictory information, traditional logic systems often fail to answer queries. A new approach tackles this persistent challenge by introducing a cost-based method for dealing with conflicting data.

The Core Innovation#

The research introduces weighted knowledge bases, where both rules and facts can be assigned numerical costs—including infinite penalties for critical constraints. Each possible interpretation of the data receives a score based on which rules and facts it violates. This cost assignment allows systems to meaningfully query databases even when they contain internal conflicts.

How It Works#

The framework defines two ways to find valid answers:

• Certain answers: Found by considering all interpretations whose total violation cost stays within an acceptable threshold
• Possible answers: Derived by examining only the lowest-cost interpretations—those that violate the fewest or least-weighted constraints

This dual approach gives users flexibility in how strictly they want results to adhere to the data's constraints.

Scope and Impact#

The researchers conducted a thorough examination of how well this method performs across different description logic systems—ranging from simpler frameworks like ELbot to more expressive ones like ALCO. Their analysis covers both combined complexity (measuring computational difficulty when both the knowledge base and query vary) and data complexity (when the query is fixed but the data changes).

Understanding these complexity tradeoffs is essential for building practical systems that can handle real-world inconsistent data at scale.