A Geometric Pattern-based Method to Build Hierarchies of Geo-Referenced Tags

To extract and structure knowledge from unstructured tag sets is a challenge for the tagging system. Although several methods have shown the ability to assign structures to the tags, their limitation still remains. Accordingly, this study investigates a method to build the place hierarchy of geo-referenced tags. In particular, the method is based on analyzing geometric patterns of footprints, derived from the tags. The objectives are to understand the structure of vernacular geography and to improve the image search by constructing tag semantics. First, the geographical regions are represented based on the large sets of the clustered collaborative tags on the place. Herein, the method to generate the chi-shape derives the footprints of the geographical regions. According to its geometric condition such as position, area, and centroid of the polygon, derived from each footprint, the place-tag hierarchies are constructed. The geographical boundaries between the place tags are established throughout the assigned footprints as well. In addition, our method enhances the accuracy of the tag hierarchies when compared to the previous research. Furthermore, the study also contributes to the development of the tag semantics and the comprehension of the vernacular geography.