Upload Dataset Card

README.md

@@ -1,33 +1,97 @@
 ---
-language:
-- en
+language: en
+license: mit
+task_categories:
+- token-classification
+task_ids:
+- named-entity-recognition
+pretty_name: PHEE test data
+tags:
+- pharmacovigilance
+- adverse-event
+- medical
+- ner
 ---
-# Dataset Card for Dataset Name
 
-<!-- Provide a quick summary of the dataset. -->
+# PHEE test data
 
-PHEE - Data identifying instances of adverse events in texts, derived from literature (PubMed)
+## Dataset Description
 
-## Dataset Details
+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 test split.
 
-### Dataset Description
+The creation of the original PHEE dataset is detailed at:
 
-<!-- Provide a longer summary of what this dataset is. -->
+> 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.
 
-Data taken from source listed below was reformatted into .CoNLL, suitable for model training
+---
+
+## 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
+
+---
 
-- **Formatted by:** christine-withers
-- **Funded by [optional]:** OpenTargets
+## Intended Use
 
-### Dataset Sources [optional]
+### Primary Use
+- Supervised NER training for biomedical NLP tasks
 
-<!-- Provide the basic links for the dataset. -->
+### 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'
+
+---
 
-- **Repository:** https://huggingface.co/datasets/sarus-tech/phee
-- **Paper:** https://arxiv.org/pdf/2210.12560
+## 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
+
+---
 
-## Uses
+## Citation
 
-<!-- Address questions around how the dataset is intended to be used. -->
+If you use this dataset, please cite the original publication:
 
-Training of NER model tagging adverse-events
+```bibtex
+@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}
+}
+```