Reconcile - Coreference Resolution Engine

Reconcile is an automatic coreference resolution system that was developed to provide a stable test-bed for researchers to implement new ideas quickly and reliably. It achieves roughly state of the art performance on many of the most common coreference resolution test sets, such as MUC-6, MUC-7, and ACE. Reconcile comes ready out of the box to train and test on these common data sets (though the data sets are not provided) as well as the ability to run on unlabeled texts. Reconcile utilizes supervised machine learning classifiers from the Weka toolkit, as well as other language processing tools such as the Berkeley Parser and Stanford Named Entity Recognition system.

The source language is Java, and it is freely available under the GPL.

Performance

MUC & B3 scores for several common coreference data sets. These experiments use the supplied model, ave_perceptron_uw, trained on the University of Wolverhampton corpus supplied with Reconcile. This model is used in the stand-alone JAR version of Reconcile as well.
MUC Score
  Recall Precision F-measure
MUC-6 67.23 65.54 66.38
MUC-7 53.27 65.69 58.83
ACE05 55.28 65.30 59.87


B3 Score
  Recall Precision F-measure
MUC-6 65.01 77.64 70.77
MUC-7 54.47 82.77 65.70
ACE05 60.05 81.49 69.14

Download: Executable JAR ZIP Tarball
Documentation: User Manual Quick Start Guide

Jar Quick Start: Run: $ java -jar reconcile-1.0.jar file1 file2 ...

Notes:

Publications:

The development of Reconcile was a collaboration between researchers from Cornell University, The University of Utah, and Lawrence Livermore National Labs. People involved include: Ves Stoyanov (JHU/Cornell), Claire Cardie (Cornell), Nathan Gilbert (Utah), Ellen Riloff (Utah), David Buttler (LLNL), and David Hysom (LLNL).

Contact: Please send questions and/or comments about Reconcile to the developer mailing list: Note: Please don't attempt to subscribe to this list, any mail sent to it will be delivered to the current Reconcile development team.