metadata

license: cc-by-sa-3.0
task_categories:
  - text-generation
  - text-classification
language:
  - 'no'
pretty_name: WIKI Paragraphs Norwegian
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.jsonl
      - split: validation
        path: validation.jsonl
      - split: test
        path: test.jsonl
      - split: validation1000
        path: validation1000.jsonl
      - split: test1000
        path: test1000.jsonl
      - split: validation100
        path: validation100.jsonl
      - split: test100
        path: test100.jsonl
      - split: pretrain
        path: pretrain.jsonl
      - split: reserve
        path: reserve.jsonl
version: 1.0.0
citation: |
  This dataset contains content from Wikipedia under CC BY-SA 3.0 license.
dataset_info:
  features:
    - name: text
      dtype: string
    - name: url
      dtype: string
    - name: paragraph_number
      dtype: int64
    - name: corrupt
      dtype: string
    - name: corrupt_level
      dtype: int64
  splits:
    - name: train
      num_examples: 1000000
    - name: validation
      num_examples: 10000
    - name: test
      num_examples: 10000
    - name: validation1000
      num_examples: 1000
    - name: test1000
      num_examples: 1000
    - name: validation100
      num_examples: 100
    - name: test100
      num_examples: 100
    - name: pretrain
      num_examples: 10000
    - name: reserve
      num_examples: 100000

WIKI Paragraphs Norwegian

A multi-split dataset for machine learning research and evaluation, containing text samples in JSON Lines format.

Features

Multiple splits for different use cases
Random shuffle with Fisher-Yates algorithm
Structured format with text and metadata
Size-varied validation/test sets (100 to 10k samples)

Splits Overview

Split Name	Samples	Typical Usage
`train`	1,000,000	Primary training data
`validation`	10,000	Standard validation
`test`	10,000	Final evaluation
`validation1000`	1,000	Quick validation
`test1000`	1,000	Rapid testing
`validation100`	100	Debugging/development
`test100`	100	Small-scale checks
`pretrain`	10,000	Pre-training phase
`reverse`	100,000	Special tasks

Total Samples: 1,132,200

License

Creative Commons Attribution-ShareAlike 3.0

This dataset inherits Wikipedia's licensing terms:

Attribution Required
ShareAlike Mandatory
Commercial Use Allowed

Usage

from datasets import load_dataset

# Load main training split
dataset = load_dataset("your-username/dataset-name", split="train")

# Access smaller validation split
val_100 = load_dataset("your-username/dataset-name", "validation100")

Data Structure

Each line contains JSON:

Copy
{
  "text": "Full text content...",
  "metadata": {
    "source": "Wikipedia", 
    "timestamp": "2023-01-01",
    "url": "https://..."
  }
}

Notes

All splits accessible via: load_dataset(repo_id, split_name) Non-standard splits (e.g., reverse) require explicit config: split="reverse" When using, include attribution: "Contains content from Wikipedia under CC BY-SA 3.0"