Improving the IR-tree in Main Memory for Spatial-Keyword Query Processing

Abstract

The exponential growth of geo-referenced data, driven by GPS-equipped mobile devices and IoT sensors, presents significant challenges for spatial-keyword (SK) query processing. Effective spatio-textual indexing is crucial for efficiently searching for these geotextual data. In this research direction, the design of in-memory spatial-keyword indices is a new and exciting research topic, which also includes optimizing processing times and reducing memory consumption. This paper addresses the design of optimized in-memory SK indices, with a focus on enhancing the in-memory IR-tree. We introduce the Sort-Tile-Recursive (STR) bulk loading algorithm during R-tree creation, resulting in a more compact index and improved performance for both index generation and query processing. Through extensive experiments using real-world datasets, we demonstrate the efficiency and effectiveness of our proposed index structures and SK query algorithms.