Keyword-Centric Community Search over Large Heterogeneous Information Networks

Community search in heterogeneous information networks (HINs) has attracted much attention in recent years and has been widely used for graph analysis works. However, existing community search studies over heterogeneous information networks ignore the importance of keywords and cannot be directly applied to the keyword-centric community search problem. To deal with these problems, we propose kKP -core, which is defined based on a densely-connected subgraph with respect to the given keywords set. A kKP -core is a maximal set of P -connected vertices in which every vertex has at least one KP -neighbor and k path instances. We further propose three algorithms to solve the keyword-centric community search problem based on kKP -core. When searching for answers, the basic algorithm Basic- kKP -core will enumerate all paths rather than only the path instances of the given meta-path P. To improve efficiency, we design an advanced algorithm AdvkKP -core using a new method of traversing the search space based on trees to accelerate the searching procedure. For online queries, we optimize the approach with a new index to handle the online queries of community search over HINs. Extensive experiments on HINs are conducted to evaluate both the effectiveness and efficiency of our proposed methods.