WindChimeRan / pytorch_multi_head_selection_re

BERT + reproduce "Joint entity recognition and relation extraction as a multi-head selection problem" for Chinese and English IE
138 stars 32 forks source link

run experiment #2

Closed antgr closed 5 years ago

antgr commented 5 years ago

Hi, thank you for this great effort. You say that it can be run using the commands: (1),(2),(3) python main.py --mode preprocessing --exp_name chinese_selection_re python main.py --mode train --exp_name chinese_selection_re python main.py --mode evaluation --exp_name chinese_selection_re

I am a bit confused. How should I reproduce the official run (CoNLL04 i think)?
(Are the above commands (1,2,3) for 2019 Language and Intelligence Challenge ?)

WindChimeRan commented 5 years ago

Hi, I update my README!

Yes, chinese_selection_re is for the competition.

If you want to try CoNLL04, please set exp_name as conll_selection_re

eg:

python main.py --mode preprocessing --exp_name conll_selection_re python main.py --mode train --exp_name conll_selection_re python main.py --mode evaluation --exp_name conll_selection_re

The hyperparameters are different from the official version, but I don't want to exactly reproduce the paper.

I am adding BERT to the model.