test-sink / README.md
librarian-bot's picture
Add dataset description
692eed4 verified
|
raw
history blame
999 Bytes
metadata
dataset_info:
  features:
    - name: id
      dtype: large_string
    - name: url
      dtype: large_string
    - name: text
      dtype: large_string
    - name: language
      dtype: large_string
    - name: token_count
      dtype: int64
    - name: dump
      dtype: large_string
    - name: page_average_lid_score
      dtype: float64
  splits:
    - name: train
      num_bytes: 3706571
      num_examples: 50
  download_size: 1790640
  dataset_size: 3706571
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This dataset contains 50 rows of text data with metadata (id, url, text, language, token count, dump, page_average_lid_score) from official/government sources (e.g., .gov.uk, .cymru URLs). All text is in Welsh (cym_Latn), with token counts ranging from 10,029 to 149,028 and a mean page_average_lid_score of 0.991. The dataset appears suitable for testing text processing pipelines or language identification tasks, given its structured format and consistent Welsh content.