wikipedia-topics / README.md
tarekziade's picture
Upload dataset
2c2dcc4
|
raw
history blame
817 Bytes
metadata
license: cc-by-sa-4.0
dataset_info:
  features:
    - name: id
      dtype: int64
    - name: title
      dtype: string
    - name: summary
      dtype: string
    - name: text
      dtype: 'null'
    - name: categories
      sequence: string
  splits:
    - name: train
      num_bytes: 35473082.703455076
      num_examples: 67570
    - name: test
      num_bytes: 3941570.2965449267
      num_examples: 7508
  download_size: 26119219
  dataset_size: 39414653
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

Creates a text/category dataset using Wikipedia.

Explores the 40 root categories and their sub-categories to collect pages that are seen only on each root category. The produced dataset provides 200 pages per category.

Author: Tarek Ziadé / Mozilla