Evaluating Name-Matching for Coreference Resolution
Computational Linguistics. Saarland University
In this paper we describe experiments aimed at improving matching techniques for the Coreference Resolution task. Combining algorithms proposed in the literature with our own solutions, we run machine learning experiments and evaluate extensively different features and feature combinations to find the best settings. These settings do not only show good performance for English data, but should also be easily adjustable to cover other languages.
Coreference Resolution, Name Matching, Minimum Edit Distance, Normalisation