Semantic frame matching, and the automatic evaluation of an information extraction system

The Frametagger is a semantic pattern matching software designed to identify the actors in short business news, developed by the Human Language Technology Group of the University of Szeged. The module is based on the semantic frames and tables developed in the NKFP 2/017/2001 project by the Research Institute for Linguistics, and extended by us later on, and is the final module of the Information Extraction Toolchain developed in Szeged. In this paper, we introduce a Benchmark algorithm as well, which is made to give a realistic insight how the development, or replacement of each module in the toolchain effects the results/accuracy of the whole Information Extraction System