We modify the code of previous works (https://github.com/ttthy/ure) to implement our experiments.
NYT: We cannoy provide NYT dataset due to copyright, please contact [Diego Marcheggiani] SemEval: https://www.aclweb.org/anthology/S10-1006/ Input format: same as sample
There are some vocabulary files needed to generate in advance. You can use the scripts
bash ure/preprocessing/run.sh python data/nyt/process.py
UREVA python -u -m ure.ureva.main --config models/ureva.yml
python -u -m ure.ureva.main --config models/ureva.yml --mode test