Research on the Feature Selection of Chinese Coreference Resolution

Coreference resolution is the process of determining whether two expressions refer to the same entity. We adopt machine learning approach to coreference resolution. Feature selection of entity is the key of coreference resolution. This paper presents analysis methods for features which are used commonly in coreference resolution, proposes two features, entity density and antecedent characteristics. Then based on the above features, employ the decision tree to finish the coreference resolution. Finally, the experimental results on the ACE data show that these two features can improve the performance of coreference resolution.