KLocator: An Ontology-Based Framework for Scenario-Driven Geographical Scope Resolution

The automatic extraction of geographical information from textual pieces of information is a challenging task that has been getting increasing attention from application and research areas that need to incorporate locationawareness in their methods and services. In this paper, we present KLocator, a novel ontology-based system for correctly identifying geographical entity references within texts and mapping them to knowledge sources, as well as determining the geographical scope of texts, namely the areas and regions to which the texts are geographically relevant. Compared to other similar approaches, KLocator has two important novelties: i) It does not utilize only background geographical information for performing the above tasks but allows the exploitation of any kind of semantic information that is explicitly or implicitly related to geographical entities in the given domain and application scenario. ii) It is highly customizable, allowing users to define and apply custom geographical resolution models that best fit to the domain(s) and expected content of the texts to be analyzed. Both these features, according to our experiments, manage to substantially improve the effectiveness of the geographical entity and scope resolution tasks, especially in scenarios where explicit geographical information is scarce. Keywords-Geographical Entity Resolution; Geographical Scope Resolution; Ontologies; Semantic Data.