After download, run the file main.py, you'll extract medical relations from the PDFs in the folder data/PDF_FOR_PARSE
Then you could search for the relations by giving a relation type or entity names, a graph containing the output relations will be shown on.
Except for the searching for relations stored in the knowledge graph, you can input "infer", then input a sentence, you would get the entities and corresponding predicted relations between the entities.
fine-tuning BioBERT on NER and relation extraction tasks referred from https://github.com/dmis-lab/biobert.