/CRCNN

Classifying Relations by Ranking with Convolutional Neural Networks

Primary LanguagePython

Classifying Relations by Ranking with Convolutional Neural Networks

Implementation of ACL 2015 Paper:
Classifying Relations by Ranking with Convolutional Neural Networks

Download SemEval 2010 Task 8 Dataset for Relation Classification

Here is the link to download this dataset: link

You will also need to download some pre-trained embeddings like GloVe.

Dependencies

tensorflow (1.3.0)
spacy
pandas
numpy
scikit-learn

Training

Update paths in model_config.yml, then start training as:

python3 -m train_crcnn

Evaluation

Once the training is finished and you have trained models in your model directory, evaluate a model as:

python3 -m test_crcnn --config_file <full path of saved .yml config file in your model directory --model_name <checkpoint prefix of the model you want to evaluate>