Unified Classification Model for Geotagging Websites

Abstract

The paper presents a novel approach to finding regional scopes (geotagging) of websites. Unlike the traditional approaches, which generally involve training a separate classification model for each class (region), the proposed method is based on training a single model which is used for all regions of the same type (e.g. cities). This approach is made possible by the usage of ”relative” features which indicate how a selected region matches up to other regions for a given website. The classification system uses a variety of features of different nature that have not been yet used together for machine-learning based regional classification of websites. The evaluation demonstrates the advantage of our ”one model per region type” method versus the traditional ”one model per region” approach. A separate experiment demonstrates the ability of the proposed classifier to successfully detect regions which were not present in the training set (which is impossible for traditional approaches).