ShannonAI / mrc-for-flat-nested-ner

Code for ACL 2020 paper `A Unified MRC Framework for Named Entity Recognition`

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Reproduce of the GENIA dataset

YuSawan opened this issue · comments

Hi, I cannot reproduce the genia dataset results for f1=83.75 in your paper; the results only reached f1=78.
If available, would you like to give your preprocessed dataset?

@YuSawan Neither do I. Have you reproduced the genia dataset results for f1=83.75 in your paper now?

Not yet. I use the genia dataset provided in Obaja's project but it can not reproduce the results.
I think this model uses different preprocessing like eliminates entities with coordinate structure (ex. Human A and B lymphocytes).

Hello! I ran your program on the genia dataset, and the result was 74. Can you give me the data set you have processed?

Please use our released data files link for GENIA.
Hyperparameters for reproducing the result can be found here.
Many thanks.

Any update of this issue? I have run the experiment with Genia dataset, but only f1 0.79 I can get. I already used your dataset and your genia.sh scripts.