Extending Graph Pattern Matching with Regular Expressions

Abstract

Graph pattern matching, which is to compute the set M(Q, G) of matches of Q in G, for the given pattern graph Q and data graph G, has been increasingly used in emerging applications e.g., social network analysis. As the matching semantic is typically defined in terms of subgraph isomorphism, two key issues are hence raised: the semantic is often too rigid to identify meaningful matches, and the problem is intractable, which calls for efficient matching methods. Motivated by these, this paper extends matching semantic with regular expressions, and investigates the top-k graph pattern matching problem. (1) We introduce regular patterns, which revise traditional pattern graphs by incorporating regular expressions; extend traditional matching semantic by allowing edge to regular path mapping. With the extension, more meaningful matches could be captured. (2) We propose a relevance function, that is defined in terms of tightness of connectivity, for ranking matches. Based on the ranking function, we introduce the top-k graph pattern matching problem, denoted by $\mathsf {TopK}$. (3) We show that $\mathsf {TopK}$ is intractable. Despite hardness, we develop an algorithm with early termination property, i.e., it finds top-k matches without identifying entire match set. (4) Using real-life and synthetic data, we experimentally verify that our top-k matching algorithms are effective, and outperform traditional counterparts.