drugscom_reviews / README.md
Zakia's picture
Updated wording in the contributions section
9e2958a
metadata
pretty_name: DrugsCom Reviews
annotations_creators:
  - no-annotation
language_creators:
  - found
language:
  - en
tags:
  - health
  - medicine
  - patient reviews
  - drug reviews
  - sentiment analysis
license:
  - cc-by-4.0
multilinguality:
  - monolingual
size_categories:
  - 100K<n<1M
source_datasets:
  - original
task_categories:
  - text-classification
task_ids:
  - sentiment-classification

Dataset Card for "DrugsCom Reviews"

Table of Contents

Dataset Description

Dataset Summary

The DrugsCom Reviews dataset is originally sourced from the UCI Machine Learning Repository. It provides patient reviews on specific drugs along with related conditions and a 10-star patient rating reflecting overall patient satisfaction. The dataset has been uploaded to Hugging Face to facilitate easier access and use by the machine learning community. It contains 161,297 instances in the training set and 53,766 instances in the test set.

Supported Tasks and Leaderboards

This dataset can be used for sentiment analysis and text classification tasks.

Languages

The text in the dataset is in English.

Dataset Structure

Data Instances

A data instance from the train split:

{
    "drugName": "Buprenex",
    "condition": "Pain",
    "review": "I have severe drug allergies especially opiates, I have been on Buprenorphine for many years now and found it to be an excellent pain reliever. I have found that at times though it is hard to find and have had to go to the patch.",
    "rating": 10,
    "date": "May 11, 2012",
    "usefulCount": 27
}

A data instance from the test split:

{
    "drugName": "Nasacort Allergy 24HR",
    "condition": "Allergic Rhinitis",
    "review": "Since I start using this product I experienced change of vision and headaches.",
    "rating": 3,
    "date": "September 8, 2015",
    "usefulCount": 27
}

plain_text

  • Size of the training dataset file: 82.314 MB
  • Size of the test dataset file: 27.414 MB
  • Total size of downloaded dataset files: 109.73 MB

A data instance consists of the following fields:

  • drugName: The name of the drug reviewed.
  • condition: The condition for which the drug was prescribed.
  • review: The text of the review by the patient.
  • rating: A patient satisfaction rating out of 10.
  • date: The date when the review was posted.
  • usefulCount: The number of users who found the review useful.

Data Fields

  • drugName: string
  • condition: string
  • review: string
  • rating: integer (0-10)
  • date: date
  • usefulCount: integer

Data Splits

The dataset is split into training and testing sets as follows:

  • train: 161,297 instances
  • test: 53,766 instances

Dataset Creation

Curation Rationale

The dataset was curated with the intention to study sentiment analysis of drug experience and the transferability of models among different domains and data sources.

Source Data

Initial Data Collection and Normalization

The dataset was collected by crawling online pharmaceutical review sites. No additional preprocessing or normalization has been conducted on the data provided in this repository; it is presented in its original form as obtained from the source.

Who are the source language producers?

The reviews were written by patients and users of the drugs.com website.

Annotations

Annotation process

No additional annotation process was followed as the data contains self-reported patient ratings.

Who are the annotators?

N/A

Personal and Sensitive Information

The dataset contains sensitive information in the form of patient drug reviews.

Considerations for Using the Data

Social Impact of Dataset

The dataset can be used to study the impact of drugs on patients, which can be beneficial for healthcare research.

Discussion of Biases

No known biases but users should consider the self-reported nature of the data.

Other Known Limitations

The dataset may not generalize well to drugs or conditions not represented in the data.

Additional Information

Dataset Curators

Curated by the UCI Machine Learning Repository.

Licensing Information

Licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0).

Citation Information

When using this dataset, please cite the original source and also the Hugging Face dataset repository:

@misc{misc_drug_review_dataset_(drugs.com)_462,
author = {Kallumadi,Surya and Grer,Felix},
title = {Drug Review Dataset (Drugs.com)},
year = {2018},
howpublished = {UCI Machine Learning Repository},
note = {DOI: https://doi.org/10.24432/C5SK5S}
}

And:

@misc{huggingface:drugscom_reviews,
title = {Drug Review Dataset (Drugs.com) - Hugging Face Version},
author = {Salod, Zakia},
year = {2023},
publisher = {Hugging Face},
howpublished = {Hugging Face Datasets Library},
url = {https://huggingface.co/datasets/Zakia/drugscom_reviews}
}

Contributions

This dataset was uploaded to Hugging Face by Zakia. Special thanks to the community for their interest and engagement. Future contributions, including discussions, issues, and improvements to the dataset card, are welcomed and greatly appreciated in the discussions section of this dataset's Hugging Face page.