Location-based keyword search: enhancing IR-tree querying

Abstract

Location-based and keyword search query has been increasing in popularity throughout the years. This type of query makes use of the location information and tagged documents to locate the top K most relevant point of interest. There exists an indexing framework, IR-Tree, that allows efficient processing of such query by combining the use of inverted file for tagged documents and R-Tree for location information. However, there is still limitation on the I/O cost for loading the inverted files when processing a query. Therefore, in this paper an enhanced implementation of the IR-Tree that incorporate B+ Tree Indexing for the inverted files will be introduce. Result on the evaluation of the enhanced implementation also shows significant improvement on the performance.