Using markov models for named entity recognition in German newspapers. Rössler, M. In Proceedings of the Workshop on Machine Learning Approaches in Computational Linguistics, pages 29–37, 2002. Citeseer.
Using markov models for named entity recognition in German newspapers [link]Website  abstract   bibtex   
This paper describes preliminary experiments for a system of named entity recognition in German newspapers. The approach is based on second order Markov Models trained on a tagged corpus. No gazetteers are used, only a list of words providing evidence is integrated. These words are extracted by statistical methods from an annotated corpus. The input basically consists of a part of speech tagged text, except the words occurring in the gained list, which replace the tags with their word form. The experiments investigate in how far such a limited approach is suitable for German and show that it provides some evidence. However, of course, it has to be enhanced.

Downloads: 0