Resources to facilitate progress in place name identification and reference resolution

This paper addresses the issue of enabling improved system performance on analyzing place names in text, especially with respect to resolving place name homonyms, and determining the unique physical location referenced by a name. The utility of reference gazetteers is discussed, and some major characteristics of three broad-coverage ones are described. Ideas are outlined for a community resource comprising one or more reference gazetteers and a variety of algorithms that take named entity tags and gazetteer entry information as input and seek to identify the "location focus/foci" of a document for use by downstream NLP system components and - via a feedback loop - by a named entity recognizer. Input from the human language technology community is sought concerning the ideas presented in the paper.