Hugging Face's logo

Datasets:
Ayomidejoe
/
NaijaMed_QA_Dataset

Tasks:
Text Classification
Feature Extraction
Question Answering
Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
1K - 10K
Tags:
medical
DOI:
Libraries:
Datasets
pandas
License:
Dataset card Viewer Files Community
2
NaijaMed_QA_Dataset / README.md
Ayomidejoe's picture
Ayomidejoe
Update README.md
2ea3d93 verified
preview code
|
raw
history blame contribute delete
3.8 kB
metadata
license: mit
task_categories:
  - text-classification
  - feature-extraction
  - question-answering
language:
  - en
tags:
  - medical
size_categories:
  - 100K<n<1M

Nigerian Healthcare Forum Q&A Datase

Dataset Summary

This dataset contains questions from Nigerians on a dedicated healthcare forum and responses provided exclusively by licensed and trained healthcare professionals. It reflects health concerns within the Nigerian context, incorporating English and local colloquialisms. All answers are reliable, as the platform restricted responses to verified healthcare professionals, ensuring the quality and credibility of the information. The dataset comprises a total of 336,681 GPT-2 tokens.

Data Collection

The data was curated from the activities on an online platform's open forum page that offered the Nigerian public free advice and answers between 2018 and 2022. It includes:

  • 1,731 questions
  • 2,639 replies/comments The data was extracted from a MySQL dump after the platform was closed.

Data Cleaning

The cleaning process involved:

  • Deduplication: Removing duplicate entries to ensure uniqueness.
  • Manual Review: Verifying the accuracy and relevance of the data.
  • Normalization: Standardizing data formats to match the quality of other relevant datasets.

Dataset Structure

Field Type Description
thread_id string The ID of the original thread
category_id string The ID of the category
category_name string The corresponding name of the category
Question string The text of the question asked
Question_token_count integer The number of GPT-2 tokens in the question text
Question_text_length integer The length of the question text (in characters)
Question_language_probability float The confidence score of the language detection algorithm
Answer string The text of the answer given
Answer_token_count integer The number of GPT-2 tokens in the answer text
Answer_text_length integer The length of the answer text (in characters)
Answer_language_probability float The confidence score of the language detection algorithm

NB: The rows are questions alongside their answers/responses/comments (multiple rows with the same thread ID). So, a single question could have more than one answers/response/comment.

How to Load the Dataset

Here's how you can load and explore the dataset using the 🤗 Datasets library:

import datasets
dataset = datasets.load_dataset('Ayomidejoe/NaijaMed_QA_Dataset')
# Display the first example
print(dataset['train'][0])

Usage

This dataset can be valuable for:

  • Natural Language Processing (NLP): Training models on Nigerian English and local colloquialisms.
  • Healthcare Analysis: Understanding prevalent health concerns in Nigeria.
  • Chatbot Training: Developing AI assistants for healthcare support in Nigerian contexts.

License

MIT

Citation

If you use this dataset in your research, please cite it as:

  title   = {Nigerian Healthcare Forum Q\&A Dataset},
  author  = {Ayomide Owoyemi}
  year    = {2024},
  publisher = {Hugging Face Datasets},
  version   = {1.0.0},
  url       = {https://huggingface.co/datasets/Ayomidejoe/NaijaMed_QA_Dataset},
}