Annotating Geographical Entities

This paper describes a study based on exploration of relations between geographical entities. We suggested a new tool for training and evaluation required by related annotation experiments. It relates to an annotator used for semi-automatic annotation, starting with the geography manual. We define fifteen types of entities: location, geo_position, geology, landform, clime, water, dimension, person, organization, URL, Timex, resource, industry, cultural, unknown with their specific subtypes. Moreover, we present the annotation conventions for three semantic relations: referential, structural and spatial, considered to be optimal operators in understanding a geographical manual. A part of the annotation is done manually, while the other part is done automatically, such as the token, lemma, part-of-speech. The study is intended to create a tool for the automatic detection of semantic relations in texts on geographic issues such as geography manuals, travel guides, geography atlases, etc., in order to help children, professors, guides, PR specialists and to be useful for tourists, generally to discover the complexity and the beauty of the nature.