A Spatial Structure Matching Algorithm for Large Spatial-Textual Datasets

Many techniques have been proposed to query and search on the massive spatial-textual data for supporting various location-based services. Traditional spatial keyword query (SKQ) and the recent spatial pattern matching (SPM), however, cannot accurately capture users' query intention on the interested objects and their spatial relations, resulting in a huge number of irrelevant results. As a result, existing techniques still cannot well satisfy users' query requirements. Therefore, this paper proposes a novel query problem-Spatial Structure Matching (SSM), which allows the user to provide a set of keyword objects and meanwhile define the constraints on both distance and direction for any two concerned objects. To answer an SSM query, this paper firstly presents a multi-way join based baseline approach, which decomposes the SSM query into the keyword matching for each individual object, edge matching for a pair of objects, and aggregation matching for a set of objects. To improve the efficiency, we further propose a sweep-line based edge matching algorithm by exploiting the geographic locations of objects to filter out the pairs of objects that cannot meet distance constraints. In addition, we build an object-connection graph using the objects that meet all constraints, and transform the SSM query into a subgraph isomorphism problem, which searches the subgraphs with similar structure as SSM query over the object-connection graph. This problem is well solved by adopting a fast subgraph matching algorithm. Experiments based on four large real-world spatial-textual datasets demonstrate that the proposed approach significantly outperforms the compared approaches by providing refined and effective results, while reducing the query processing time by at least 3 times. © 2022, Science Press. All right reserved.