Report on novel spatial-keyword search techniques

Abstract

Nowadays an increasing amount of web-accessible information on spatial objects becomes available to the public every day. Apart from the spatial location of an object (e.g., a point of interest), additional descriptive information typically includes textual description as well as various ratings, often user generated. Modern applications employ spatio-textual queries, which take into account both the spatial location of an object and its textual similarity to retrieve the most relevant objects. However, existing applications provide a limited functionality to the users. For example, several meaningful queries cannot be expressed by existing approaches and motivate our novel prototype system. In the first chapter, we address this limitation by supporting ranked retrieval of objects of interest by taking into account the quality of facilities in their vicinity, but also their textual similarity to user defined keywords. In the second chapter, we analyze the properties of geotagged photos of Flickr, and propose novel location-aware tag recommendation methods. Both of the above techniques are novel spatial-keyword search methods.