Named Entity Recognition for the Mainland Scandinavian Languages.

In this paper we discuss the results of the Nomen Nescio Named Entity Recognition project, a joint effort for the mainland Scandinavian languages—Norwegian, Swedish, and Danish. Fiver research groups have been involved, and developed NE recognizers using rule-based as well as statistical methods. We focus particularly on the choice of semantic categories and the problems regarding metonymy and semantic polysemy. Furthermore, we discuss the extent to which different approaches to these problems have different effects on the different types of systems, and look at two strategies, which we call Function over Form, and Form over Function. [ABSTRACT FROM AUTHOR]