Ontology-based recommendation for points of interest retrieved from multiple data sources

Abstract

Introduction of powerful mobile devices and increasing availability of online services make it possible to develop a wide range of mobile applications. Making recommendations to the users on their mobile devices based on their location is a wellknown application area of location based services. In this work we introduce an ontology based approach to find reasonable recommendations for sites (Points of Interest) like restaurants, hotels, and touristic places. We extend an existing OWL ontology in order to keep semantic relationships between different site types. Our application populates this ontology with site instances collected from several data sources, namely Google Maps, GeoNames, DBpedia and a local database. During this integration process, solutions for ontology mapping, site categorization, and duplicate site detection are developed. The ontology is then used to make recommendations on a mobile augmented reality application based on user's inputs on his device.

Detecting Places of Interest Using Social Media

“…However, both the name and location of duplicate entries may be slightly different. Therefore, we have used a heuristic based on the approach from [9] to detect and remove duplicates: first, places are indicated as duplicates when they are located closer than 5 meters of each other. Second, to detect additional duplicates of a given place p all neighboring places of the same type in a range of 100 meter were selected as candidate duplicates.…”

Using social media to find places of interest

“…To automatically estimate the semantic type of places, some authors have suggested to classify places based on their name [15], the content of webpages which contain the place name [3] or both [16]. However, only a few broad place types were considered in these works and the proposed methods have only been tested on small test sets ranging from 59 to 1160 places.…”

“…However, both the name and location of duplicate entries may be slightly different. Therefore, we have used a heuristic based on the approach from [15] to detect and remove duplicates: first, places are indicated as duplicates when they are located closer than 5 meters to each other. Second, to detect additional duplicates of a given place p all neighboring places of the same type in a range of 100 meter were selected as candidate duplicates.…”

“…Finally, given the tags associated to each place type, similarities between place types can be estimated to enrich place type ontologies [15]. For example, using the Jensen-Shannon divergence (JSD) between the tag probabilities of two place types, we can determine that restaurants are semantically most related to pubs (JSD of 255 748), and that restaurants are more related to shops (JSD of 801 100) than to museums (JSD of 941 575).…”

A Spatial-temporal POI Data Model for Implementing Location-based Services

“…The location-based service is the service provided based on the location information of user (Ozdikis et al, 2011) and map data as the background to show the position visually and the information regarding the POI (point of interest) showing the position of main objects are necessary (Rae et al, 2012).…”