Datasets:
language: en
license: mit
task_categories:
- token-classification
task_ids:
- named-entity-recognition
pretty_name: PHEE train data
tags:
- pharmacovigilance
- adverse-event
- medical
- ner
PHEE train data
Dataset Description
This dataset contains sentences derived from medical case report abstracts curated for adverse events. Split data and CoNLL formatting allows for the training of language models, for named entity recognition. The dataset includes entity annotations or labels. This subsect is the train split.
The creation of the original PHEE dataset is detailed at:
Sun, Z., Li, J., Pergola, G., Wallace, B. C., John, B., Greene, N., Kim, J., & He, Y. (2022). PHEE: A dataset for pharmacovigilance event extraction from text. arXiv preprint arXiv:2210.12560. https://arxiv.org/pdf/2210.12560.
Source Data
The port of the original PHEE dataset used for our purposes is detailed here:
Original source repository:
https://huggingface.co/datasets/sarus-tech/phee
Intended Use
Primary Use
- Supervised NER training for biomedical NLP tasks
Not Intended For
- Clinical or patient-level decision making
Dataset Structure
- Language: English
- Splits: Train / Test / Validation
- Features: Text field, BIO label
- Labels: Adev ~ 'Adverse Event'
Preprocessing
- Sentence-level segmentation is enforced
- Annotations carried out by 15 annotators in data's original creation
- Present dataset split into train / test / val
- Present dataset labeled in the IOB CoNLL format
Limitations
- Relatively small corpus size compared to large-scale pretraining datasets
- Specific to medical case report abstracts only
Ethical Considerations
- All content originates from publicly available, open-access scientific datasets
- No personal, clinical, or identifiable patient information is included
Citation
If you use this dataset, please cite the original publication:
@article{sun2022phee,
title = {PHEE: A dataset for pharmacovigilance event extraction from text},
author = {Sun, Z., Li, J., Pergola, G., Wallace, B. C., John, B., Greene, N., Kim, J., & He, Y.},
journal = {arXiv},
year = {2022},
doi = {preprint arXiv:2210.12560}
}