Date of Award
Doctor of Philosophy (PhD)
ScienceWeb is a semantic web system that collects information about a research community and allows users to ask qualitative and quantitative questions related to that information using a reasoning engine. The more complete the knowledge base is, the more helpful answers the system will provide. As the size of knowledge base increases, scalability becomes a challenge for the reasoning system. As users make changes to the knowledge base and/or new information is collected, providing fast enough response time (ranging from seconds to a few minutes) is one of the core challenges for the reasoning system.
There are two basic inference methods commonly used in first order logic: forward chaining and backward chaining. As a general rule, forward chaining is a good method for a static knowledge base and backward chaining is good for the more dynamic cases. The goal of this thesis was to design a hybrid reasoning architecture and develop a scalable reasoning system whose efficiency is able to meet the interaction requirements in a ScienceWeb system when facing a large and evolving knowledge base.
Interposing a backward chaining reasoner between an evolving knowledge base and a query manager with support of "trust" yields an architecture that can support reasoning in the face of frequent changes. An optimized query-answering algorithm, an optimized backward chaining algorithm and a trust-based hybrid reasoning algorithm are three key algorithms in such an architecture. Collectively, these three algorithms are significant contributions to the field of backward chaining reasoners over ontologies.
I explored the idea of "trust" in the trust-based hybrid reasoning algorithm, where each change to the knowledge base is analyzed as to what subset of the knowledge base is impacted by the change and could therefore contribute to incorrect inferences. I adopted greedy ordering and deferring joins in optimized query-answering algorithm. I introduced four optimizations in the algorithm for backward chaining. These optimizations are: 1) the implementation of the selection function, 2) the upgraded substitute function, 3) the application of OLDT and 4) solving of the "owl: sameAs" problem.
I evaluated our optimization techniques by comparing the results with and without optimization techniques. I evaluated our optimized query answering algorithm by comparing to a traditional backward-chaining reasoner. I evaluated our trust-based hybrid reasoning algorithm by comparing the performance of a forward chaining algorithm to that of a pure backward chaining algorithm. The evaluation results have shown that the hybrid reasoning architecture with the scalable reasoning system is able to support scalable reasoning of ScienceWeb to answer qualitative questions effectively when facing both a fixed knowledge base and an evolving knowledge base.
"Scalable Reasoning for Knowledge Bases Subject to Changes"
(2014). Doctor of Philosophy (PhD), Dissertation, Computer Science, Old Dominion University, DOI: 10.25777/sh7k-7a32