Quickly locating POIs in large datasets from descriptions based on improved address matching and compact qualitative representations

Locating points of interest (POIs) from descriptions can support intelligent location-based services. Available research achieves it through address matching and spatial reasoning. However, semantic characteristics and spatial proximities of address fields are usually neglected in address matching; current applications of spatial reasoning represent qualitative spatial relations in semantic networks for efficient queries, but they do not yet scale to large datasets for qualitative direction reasoning due to massive qualitative direction relations between objects; moreover, spatial reasoning on various quantitative distances should be optimized. This study proposes a method that improves the accuracy of address matching by combining multiple similarities and enables quick spatial reasoning through the faster relation retrieval of compact qualitative direction representations implemented on global equal latitude and longitude grids (ELLGs) and the ELLG-based quantitative calculations. The proposed method has been verified by two real-world datasets and proven to be efficient and accurate when locating POIs in large POI datasets from descriptions.