esmailza / chemical-gene-chemical-gene-relation-extraction-with-gnn

The model identifies chemical components and genes named entities and extracts the relations of the chemical-gene pair jointly. It utilizes the BioBERT model in the named entity recognition and the graphs neural networks for the RE subtasks.

Chemical-Gene-Relation-Extraction-with-GNN

The model identifies chemical components and genes named entities and extracts the relations of the chemical-gene pair jointly. It utilizes the BioBERT model for word embedding and uses the GNN model for named recognition and RE subtasks.

Model Architecture

• Word Embedding using BioBERT
• Graph Attention Networks
  • Calculate Attention Weights
  • Update Vector Representations
• Relation Extraction and Tagging
  • Chemical Tagger
  • Gene Tagger

Train the model

python trainBioCreative.py

Requirements

• python 3.7
• torch 1.3
• tqdm
• transformers
• numpy

Citation

@inproceedings{esmail2022chemical,
  title={Chemical-Gene Relation Extraction with Graph Neural Networks and BERT Encoder},
  author={Esmail Zadeh Nojoo Kambar, Mina and Esmaeilzadeh, Armin and Taghva, Kazem},
  booktitle={The International Conference on Innovations in Computing Research},
  pages={166--179},
  year={2022},
  organization={Springer}
}

chemical-gene-chemical-gene-relation-extraction-with-gnn's Issues

Architecture

Hello,
I'm confused it seems that your gnn layers are not used anywhere in GNN class? If they actually are used somewhere can you point out where ? Thank you