Datasets:

Languages:
English
Multilinguality:
monolingual
Size Categories:
100K<n<1M
Language Creators:
found
expert-generated
Source Datasets:
original
ArXiv:
DOI:
License:
cogtext / README.md
morteza's picture
Update README.md
4ff105b
metadata
pretty_name: CogText PubMed Abstracts
license:
  - cc-by-4.0
language:
  - en
multilinguality:
  - monolingual
task_categories:
  - text-classification
task_ids:
  - topic-classification
  - semantic-similarity-classification
size_categories:
  - 100K<n<1M
paperswithcode_id: linking-theories-and-methods-in-cognitive
inference: false
model-index:
  - name: cogtext-pubmed
    results: []
source_datasets:
  - original
language_creators:
  - found
  - expert-generated
configs:
  - config_name: abstracts (2023)
    data_files: pubmed/abstracts2023.csv.gz
  - config_name: abstracts (2021)
    data_files: pubmed/abstracts2021.csv.gz
tags:
  - Cognitive Control
  - PubMed

Dataset Card for CogText PubMed Abstracts

Table of Contents

Dataset Description

The CogText dataset is a curated collection of abstracts about cognitive tasks and constructs from PubMed. This dataset contains the original abstracts and their corresponding embeddings. Please visit CogText on GitHub for the details and codes.

Dataset Summary

The 2021 dataset, collected in December 2021, contains 385,705 distinct scientific articles, featuring their title, abstract, relevant metadata, and embeddings. The articles were specifically selected for their relevance to cognitive control constructs and associated tasks.

Supported Tasks and Leaderboards

Topic Modeling, Text Embedding

Languages

English

Dataset Structure

Data Instances

522,972 scientific articles, of which 385,705 are unique.

Data Fields

The CSV files contain the following fields:

Field Description
index (int) Index of the article in the current dataset
pmid (int) PubMed ID
doi (str) Digital Object Identifier
year (int) Year of publication (yyyy format)
journal_title (str) Title of the journal
journal_iso_abbreviation (str) ISO abbreviation of the journal
title (str) Title of the article
abstract (str) Abstract of the article
category (enum) Category of the article, either "CognitiveTask" or "CognitiveConstruct"
label (enum) Label of the article, which refers to the class labels in the ontologies/efo.owl ontology
original_index (int) Index of the article in the full dataset (see pubmed/abstracts.csv.gz)

Data Splits

Dataset Description
pubmed/abstracts.csv.gz Full dataset
pubmed/abstracts20pct.csv.gz 20% of the dataset (stratified random sample by label)
gpt3/abstracts_gp3ada.nc GPT-3 embeddings of the entire dataset in XArray/CDF4 format, indexed by pmid

Dataset Creation

Curation Rationale

[Needs More Information]

Source Data

Initial Data Collection and Normalization

[Needs More Information]

Annotations

Annotation process

[Needs More Information]

Personal and Sensitive Information

[Needs More Information]

Considerations for Using the Data

Social Impact of Dataset

[Needs More Information]

Discussion of Biases

[Needs More Information]

Other Known Limitations

[Needs More Information]

Additional Information

Dataset Curators

[Needs More Information]

Licensing Information

[Needs More Information]

Acknowledgments

This research was supported by the Luxembourg National Research Fund (ATTRACT/2016/ID/11242114/DIGILEARN and INTER Mobility/2017-2/ID/11765868/ULALA).

Citation Information

To cite the paper use the following entry:

@misc{cogtext2022,
  author = {Morteza Ansarinia and
            Paul Schrater and
            Pedro Cardoso-Leite},
  title = {Linking Theories and Methods in Cognitive Sciences via Joint Embedding of the Scientific Literature: The Example of Cognitive Control},
  year = {2022},
  url = {https://arxiv.org/abs/2203.11016}
}