Corneferencer

Corneferencer is a neural network based tool for performing Coreference Resolution (CR).

Principal developer: Bartłomiej Nitoń
Authors: Bartłomiej Nitoń, Paweł Morawiecki

Download

You may check out the source code from git repository.

Models

• 1147 features (word embedding vectors size 50)

• 6647 features (word embedding vectors size 300)

• 1190 features (word embedding vectors size 50)

• Siamese Networks model (word embedding vectors size 50)

• 1190 features (5 times more negative examples used for training, word embedding vectors size 50)

• word embedding vectors

Evaluation

• test data set (TEI)

• training data set (TEI)