Resources for Place Name Analysis
Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC 2004)
Abstract
We present a new resource for annotating and visualizing the meaning of place names in natural language text, along with insights gained from analysis of manual annotations. The work addresses the issue of place name (toponym) meaning resolution in text, moving beyond simple named entity recognition to address the problem of grounding textual references, i.e., making a connection between the references and the real-world entities that they denote. The name, "San Francisco," for example, can be mapped to more than 900 distinct place entities that differ in terms of location and/or type, according to commonly available databases of named geographical entities, called gazetteers. Gazetteers serve as knowledge bases that can be exploited to support the analysis and disambiguation of named places, and the grounding of textual references in real-world entities. This process of grounding text in gazetteers offers a way of normalizing the meaning of the place name references that are found in the gazetteers, and essentially subsumes the text analysis process of determining when instances of a given name are being used to refer to the same or to different entities.