Multilingual Mention Detection for Coreference Resolution

This paper proposes a novel algorithm for multilingual mention detection: we extract mentions from parse trees via kernelbased SVM learning. Our approach allows for straightforward mention detection for any language where (not necessary perfect) parsing resources are available, without any complex language-specific rule engineering. We also investigate possibilities for incorporating automatically acquired mentions into an end-to-end coreference resolution system. We evaluate our approach on the Arabic and Chinese portions of the CoNLL-2012 dataset, showing a significant improvement over the system with the baseline mention detection.