bvanaken

/

clinical-assertion-negation-bert

like 28

1. Update label2id and id2label mapping
2. Updated model inference with pipeline

really cool!

I'm assuming the fine-tuning dataset isn't present on hf.co (biomedical datasets are sensitive so sometimes hard to move around) otherwise you could link the model to the dataset using datasets: in the metadata

Thanks, Kamal. Looks great!

@julien-c : Yes, right. The dataset behind the model is based on MIMIC-III and requires registration (and training) at PhysioNet, so unfortunately we currently cannot share it on Hugging Face.

Got it! For context, we are in the process of building more tools/workflows around access to datasets, which we hope will be useful to the community – but I recognize that biomedical datasets are particularly sensitive!