Busca escalável de região geográfica similar utilizando BM25.

Abstract

Information retrieval applications are increasingly robust, versatile and serving the most diverse purposes. In the area of Geographic Information Retrieval (GIR), although many approaches have been proposed to deal with several problems, some problems still remain insufficiently explored, such as approaches that allow searching for regions that are similar to a region of interest in a fast and scalable way . Given a spatial region and a query region, a similar region search aims to find the K regions most similar to the query region in the spatial region. This work presents a similarity search approach in maps using a textual similarity algorithm with stored and indexed data. For this, an algorithm was developed based on the use of indexes in ElasticSearch to store geographic data. Queries are textual, using the concept of inverted index together with the BM25 similarity algorithm. In this algorithm, the Points of Interest (POI) information present in the regions of the maps are converted into textual data that are the basis for carrying out queries. The proposed approach is based on a case study using data from large cities obtained through OpenStreetMaps.