search

aistairc / DeepEventMine

DeepEventMine: End-to-end Neural Nested Event Extraction from Biomedical Texts
Apache License 2.0
95 stars 20 forks source link

How to train DeepEventMine on my own dataset? #17

Open YaoXinZhi opened 1 year ago

YaoXinZhi commented 1 year ago

Hi Alistair.

Thank you for your exciting work on the BioEE task.

I was wondering if DeepEventMine could be easily trained on my own dataset and used for the annotation of new text.

Because I noticed that "event structures files" are not easy to prepare.

Just like mentioned in the 'run/prepare-cg.sh'

`

  1. EVENT STRUCTURES echo "-----------------------------" echo "3. Download processed event structures"

    Our processed event structures data FILEID=1qzjY38RdnCvomVAD3XMk_J4pi1gIraA5

    gdown https://drive.google.com/uc?id=$FILEID tar -xzvf cg.tar.gz rm *.tar.gz

`

Do you have any suggestions for the preparation of this file?

You may have heard about Active gene annotation corpus (AGAC), a biological event annotation corpus developed in recent years, and we would like to use DeepEventMine for AGAC training and annotation.

Looking forward to your reply!

best wish, xinzhi.

foodaiproject commented 1 year ago

I have a same problem, did you solve any problem?

trieuhl commented 1 year ago

Please use this code in the branch "train" to train the model https://github.com/aistairc/DeepEventMine/tree/train