This is the source code for our paper "Combining knowledge and context representations for chemical-disease relation extraction".

Keras> 2.0.1
tensorflow> 1.0
python 2.7

• cdr_model.py: The NAM model for CDR extraction.
• customize_layer.py: The attention layer used in NAM.
• cnn.py: A simple demo for CDR extraction.

We also release our code for the paper "Chemical-induced Disease Relation Extraction with Dependency In-formation and Prior Knowledge" The model is similar with the NAM, however, we use the CNN to capture the semantic dependency representations (SDR).
Then, an attention mechanism is employed to learn the importance/weight of the SDR.

• can_model: The CAN model for CDR extraction.