This package contains a Coreference Resolver (presented at EMNLP 2008), along with a collection of coreference related features.
These features include gender and number match, WordNet relations including synonym, hypernym, and antonym, and ACE entity types (e.g. semantic classes such as person, organization, and geopolitical entity). These features also include an anaphoricity classifier trained using machine learning techniques.
This collection of features is a key ingredient in the performance of the included coreference classifier.
To compile these features and build the coreference classifier from source, you will need LBJ version 2.2.0 or higher (e.g. LBJ2.jar and LBJ2Library.jar) and the LBJ Part-Of-Speech tagger version 1.3.0 or higher (LBJPOS.jar). To train the coreference classifier, you will need annotated training data such as the LDC's ACE 2004 corpus.