InMedData/Cardio_v2

Dataset Card

This dataset consists of abstracts from heart-related papers collected from PubMed. It can be used for pre-training a language model specialized in cardiology. The dataset was collected through the PubMed API, based on the names of heart-related journals and a glossary of cardiology terms.

Dataset

Data Sources

• Pubmed: PubMed is a database that provides abstracts of research papers related to life sciences, biomedical fields, health psychology, and health and welfare. Among these, we have collected abstracts of papers related to the heart.
• Wikipedia: Wikipedia is an internet encyclopedia that anyone can edit and is maintained through collaboration.

Keywords Sources

• Scimago Journal & Country Rank : We used a list of cardiology-related journals provided by SJR as keywords for data collection.
• National Institutes of Health : We used a glossary provided by NIH as keywords for data collection.
• The Texas Heart Institute : We used a glossary provided by Texas Heart Institute as keywords for data collection.
• Aiken Physicians Alliance : We used a glossary provided by Aiken Physicians Alliance as keywords for data collection.

Dataset Field

Field	Data Type	Description
title	string	The title of the paper.
abst	string	The abstract of the paper.

Dataset Structure

DatasetDict({
    train: Dataset({
        features: ['title', 'abst'],
        num_rows: 2761083
    })
})

Use

from datasets import load_dataset

dataset = load_dataset("InMedData/Cardio_v2")

Dataset Contact

khs1220@inmed-data.com