metadata
language:
- tr
dataset_info:
features:
- name: text
dtype: string
- name: url
dtype: string
- name: tokens
dtype: int64
- name: length
dtype: int64
splits:
- name: train
num_bytes: 4891677519
num_examples: 1845941
download_size: 2744928545
dataset_size: 4891677519
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
News in the Turkish language. There are a total of 1.8M news articles and 2.1B words. Conversion to the number of tokens depends on the tokenizer used.
- Collected from Common Crawl.
- Filtered and deduplicated using Datatrove.