training nlp models that can perform multiple tasks with the same set of representations.
pre-trained models are currently available that multitask over eight clinical note tasks.
This codebase can be utilized to replicate results for our paper. See the Replication section for details.
Install with
pip install https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/releases/v0.2.0/en_core_sci_sm-0.2.0.tar.gz
pip install git+https://github.com/AndriyMulyar/multitasking_transformers
Examples are available for training, evaluation and text prediction.
Running the script predict_ner.py will automatically download a pre-trained clinical note multi-tasking model, run the model through a de-identified clinical note snippet and display the entity tags in your browser with displacy.
See the directory /examples/experiment_replication.
https://arxiv.org/abs/2004.10220
Implementation, development and training in this project were supported by funding from the McInnes NLP Lab at Virginia Commonwealth University.