Home

Awesome

OpenRE

The source code of Relational Siamese Network (EMNLP2019)<br> Future up-to-date versions will be available in: https://github.com/thunlp/RSN

Environment

You only need TensorFlow(>=1.12.0) to run this code.<br> conda create -n rsn python=3.6<br> pip install networkx python-louvain tensorflow-gpu==1.12<br>

Data

You can download data by:<br> bash download.sh<br> The data is already preprocessed from the original FewRel and Glove data.

Running CNN

A standard CNN for supervised RE can be trained by<br> cd CNN<br> python train_CNN.py<br>

Running RSN

The Relational Siamese Network for OpenRE can be trained by<br> cd RSN<br> python train_RSN.py<br> By default it will be trained as a semi-supervised RSN.<br>

A supervised RSN can be trained by<br> python train_RSN.py --trainset_loss_type cross --testset_loss_type none<br>

Future Work

The FewRel-distant dataset and more models will be released in the future.<br>