We release the pre-trained model files and corresponding vocabulary for three benchmark datasets which leveraged in our paper. For convenience, we also upload the original data for Newsela, WikiSmall and WikiLarge, their corresponding SDP graph files and the preprocessed data through NER replace. You can only employ these data for research purposes. Any use of these data for commercial profit is not allowed.