metadata

license: apache-2.0
task_categories:
  - text-classification
language:
  - nb
  - 'no'
  - nn
size_categories:
  - 100K<n<1M
configs:
  - config_name: default
    data_files:
      - split: train
        path: train.jsonl
      - split: train_a
        path: trainA.jsonl
      - split: train_b
        path: trainB.jsonl
      - split: train_nordic
        path: train_nordic.jsonl
      - split: train_cleaned
        path: train_cleaned.jsonl
      - split: dev
        path: dev.jsonl
      - split: dev_nordic
        path: dev_nordic.jsonl
      - split: test
        path: test.jsonl
      - split: test_nordiv
        path: test_nordic.jsonl

Dataset Card for Bokmål-Nynorsk Language Detection (main_train_split)

Dataset Summary

This dataset is intended for language detection for Bokmål to Nynorsk and vice versa. It contains 800,000 sentence pairs, sourced from Språkbanken and pruned to avoid overlap with the NorBench dataset. The data comes from translations of news text from Norsk telegrambyrå (NTB), performed by Nynorsk pressekontor (NPK). In addition the dev and test set has 1000 entries.

Data Collection

Period: February 2011 to December 2022
Source: Omsetjingsminne Nynorsk Pressekontor - Språkbanken
Size: 800,000 sentence pairs
Format: JSON-lines (with text , language fields)

Processing Steps

Pruned to avoid overlap with NorBench
Deduplicated
Shuffled with a fixed seed (42)

Usage

Intended for training Bokmål-Nynorsk detection models. For more details, refer to the repository where the dataset preparation script and the actual dataset reside.