Toponym recognition in custom-made map titles
ISBN 978-85-88783-11-9
Authors
1Domingues, C.; 2Eshkol-taravella, I.
1IGN/LABORATOIRE COGIT Email: catherine.domingues@ign.fr
2UNIVERSITÉ D'ORLÉANS/LLL Email: iris.eshkol@univ-orleans.fr
Abstract
The titles of customized topographic maps constitute a specific corpus which is characterized by a very significant number of place names and spelling variations. This paper is about identifying toponyms in these titles. The toponym tracking is based on gazetteers as well as light parsing according to patterns. The method used broadens the definition of the top-onym to include the nature of the corpus and the data in it. It consists of seven successive stages where both the extralinguistic context - in this case toponym georeferencing - and the linguistic context are taken into account. Mistakes in tagging are analyzed from the corpus characteristics and the results of each step tagging are evaluated (recall, precision, F-measure). Different conclusions can be suggested: i) toponym recognition in web cor-pora should take into account spelling changes, ii) toponym recognition cannot be limited to gazetteer proper nouns, iii) the notion of subjective toponym is relevant in this specific corpus, and could be considered with reference to the customization of maps.
Keywords
Toponym; Subjective toponym; Web corpus