A planning-based graph matching algorithm for knowledge structure retrieval

This paper presents a novel and efficient graph matching algorithm knowledge structure retrieval. This new algorithm is based on the fundamental approach of Constrained Resource Planning (CRP), which is a powerful tool for solving planning and scheduling problems using a resource management focus. Useful actions in Concurrent Engineering systems, such as knowledge structure matching, retrieval and manipulation, could be effectively handled by this algorithm. Since the objects and relations of a knowledge base are treated as semantic networks, production rules, or frames represented in the form of directed, labeled graphs, the algorithm is suitable to any operations on knowledge bases that require the functions of Sub-Graph Isomorphism (SGI). By fully integrating the principles of resource sharing and utilization, the CRP-based SGI algorithm has been analyzed to achieve a near-linear complexity and has actually demonstrated this performance for a wide range of practical knowledge structure matching or randomly generated problems, when the input graphs possess the characteristics of local distinguishability. The empirical investigation and the resulting data demonstrate the quantitative and qualitative impact of label set size and node degree bound in the input graphs on time performance. The experimental results validate the analyses of algorithm performance. This confirmation boosts confidence in the CRP-SGI algorithm not only for solving a broad scope of practical applications (well beyond knowledge structure matching) but also for assisting the SGI problem solvers to predict the time and space requirements for their problems.