A High Quality Partial Parser for Annotating German Text Corpora

Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC 2004)

Abstract

In this paper, a two-stage partial parser for untagged German sentences is presented. In the first stage, the sentence is segmented into better parsable units according to the Topological Field Model. In the second stage, {\em minimal phrases\/} of NPs, DPs and PPs as well as nominal multiword units are identified in each of the recognized fields. In this paper, we discuss the results of the second stage. We evaluated 500 parsed sentences of a newspaper corpus. The achieved recall and precision rates are better than the ones of comparable systems as reported in literature so far.