license: apache-2.0
language:
- tr
task_categories:
- question-answering
- text-classification
- text-generation
- text-retrieval
tags:
- medical
- text
size_categories:
- n<1K
dataset_info:
features:
- name: category
dtype: string
- name: topic
dtype: string
- name: text
dtype: string
- name: num_tokens
dtype: int64
splits:
- name: train
num_bytes: 14981996
num_examples: 917
download_size: 5518304
dataset_size: 14981996
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Dataset Card for MedData_tr-1
This dataset has 917 instances and 5227389 tokens in total
Dataset Details
Dataset Description
- Language(s) (NLP): Turkish
- License: APACHE 2.0
Dataset Sources
Memorial Health Library : https://www.memorial.com.tr/saglik-kutuphanesi
Dataset Structure
category : The data was split into 3 categories
- Tanı ve Testler (Diagnoses and Tests)
- Hastalıklar (Diseases)
- Tedavi Yöntemleri (Treatment Methods)
topic : The topic of the text content
text : Full text
num_tokens : Token count of the full text
Dataset Creation
Curation Rationale
This dataset was created to increase the Turkish medical text data in HuggingFace Datasets library.
Source Data
Memorial is a hospital network based in Turkey. Their website provides a health library, which the contents were written by doctors who are experts in their fields.
Data Collection and Processing
The contents were scraped using Python's BeautifulSoup library.
Annotations
Each text in the dataset was tokenized and counted afterwards.
Annotation process
Tokenization was done using Tiktoken's encoding cl100k_base, used by gpt-4-turbo, gpt-4, gpt-3.5-turbo, etc.
Personal and Sensitive Information
This data does not contain ant personal, sensitive or private information.
Dataset Card Authors
Zeynep Cahan