CLIN 2005 Abstracts
  • Is Groningen referring to the city or the province? Geographic Named Entity Disambiguation
    Geoffrey Andogah (Humanitics Computing, University of Groningen
    Geographic named entities (GNEs) are ambiguous in two ways - a name may refer to various locations (i.e. Groningen, the Netherlands or Groningen, Germany), and a location may have many names (Holland and the Netherlands). In addition to the first type ambiguity, a name may have various senses (i.e. Groningen can refer to a city and a province) and may also refer to non-geographic entities (people and organizations). In this work we are concerned with "one name to many location" and "one name to many senses" ambiguity resolution. To resolve the first type of ambiguity, we determine in which geographical hierarchy a name can be placed (i.e. Groningen, the Netherlands, Europe vs. Groningen, Baden-Wurttemburg, Germany, Europe, ...). We assign a hierarchy on the basis of other (non-ambiguous) GNEs in context, and default locations (where population size can be used to determine the default location). The sense ambiguity is resolved on the basis of key-phrases in context (province, city), and the sense of other GNEs in context. We report on the result of preliminary experiment to test the hypothesis.