Firstly, achieving ACE2005 dataset with scripts provided by two-are-better-than-one. Then, getting the final data by running ace2005.sh

./ace2005.sh ace2005_folder

Note that we adopt bert-base-uncased tokenizer as the default. You can try other tokenizers by modifying ace2005.sh.

Firstly, downloading SciERC dataset from sciIE. Then, getting the final data by running scierc.sh

./scierc.sh scierc_folder

Note that we adopt bert-base-uncased tokenizer as the default. You can try other tokenizers by modifying scierc.sh.

We provide demo samples in the folder demo/.