Knowledge graph revision in the context of unknown knowledge

Area of Science:

• Artificial Intelligence
• Knowledge Representation and Reasoning

Background:

• Knowledge graphs are crucial for AI, requiring continuous updates as new information emerges or standards change.
• Updating knowledge graphs can be challenging in environments with limited or no initial knowledge, such as proprietary software systems.

Purpose of the Study:

• To investigate methods for updating knowledge graphs when initial knowledge is unavailable.
• To determine if Dalal revision operators can facilitate knowledge graph updates under such constraints.

Main Methods:

• Proved that finding optimal solutions for knowledge graph updates without prior knowledge is a strongly NP-complete problem.
• Developed two novel algorithms, Flaccid_search and Tight_search, designed for knowledge graph revision in knowledge-scarce environments.

Main Results:

• Demonstrated that the problem of updating knowledge graphs without prior knowledge is computationally complex (strongly NP-complete).
• Showcased the efficacy of both Flaccid_search and Tight_search algorithms in finding desired results for knowledge graph updates.
• Validated the applicability of Dalal revision operators in scenarios lacking initial knowledge.

Conclusions:

• It is possible to update knowledge graphs even without complete prior knowledge, utilizing specific revision operators.
• The proposed Flaccid_search and Tight_search algorithms offer viable solutions for knowledge graph updates in challenging, knowledge-limited scenarios.
• This research advances the field of knowledge representation and reasoning in artificial intelligence by addressing a critical gap in update methodologies.