albinandersson's picture
Update README.md
991de6b verified
metadata
license: cc-by-nc-sa-4.0
language:
  - sv
train_size: 300075
test_size: 5000
content_types:
  - biographical
  - geographical
  - scientific
  - historical
  - cultural
  - technical
pretty_name: Nordisk Familjebok Headword Extraction Dataset
task_categories:
  - text-classification
size_categories:
  - 100K<n<1M
libraries:
  - datasets
  - pandas
  - polars
  - croissant
modalities:
  - text
formats:
  - jsonl
date_published: '2024-12-09'
version: 1.0.0
cite_as: >-
  @inproceedings{andersson2026atlas, author    = {Albin Andersson and Salam
  Jonasson and Fredrik Wastring and Pierre Nugues}, title     = {ATLAS: Article
  Tracking, Linking, and Analysis of Swedish Encyclopedias}, year      = {2026},
  pages     = {3566--3576}, doi       = {10.63317/5idqekcpsxyg} }
croissant:
  '@context':
    '@vocab': http://mlcommons.org/croissant/
    schema: https://schema.org/
    citeAs: http://mlcommons.org/croissant/citeAs
  '@type': Dataset
  name: Nordisk Familjebok Headword Extraction Dataset
  description: >-
    A dataset extracted from Nordisk Familjebok used for training a headword
    extraction model.
  url: https://huggingface.co/datasets/albinandersson/nf-headword-extraction
  version: 1.0.0
  datePublished: '2024-12-09'
  citeAs: >-
    @inproceedings{andersson2026atlas, author    = {Albin Andersson and Salam
    Jonasson and Fredrik Wastring and Pierre Nugues}, title     = {ATLAS:
    Article Tracking, Linking, and Analysis of Swedish Encyclopedias}, year     
    = {2026}, pages     = {3566--3576}, doi       = {10.63317/5idqekcpsxyg} }
  license: https://creativecommons.org/licenses/by-nc-sa/4.0/
  creator:
    '@type': Organization
    name: Albin Andersson, Salam Jonasson, Fredrik Wastring, Pierre Nugues

Dataset Name: Nordisk Familjebok Headword Extraction Dataset

Description

This dataset contains pairs of extracted text and their corresponding headwords from all editions of Nordisk Familjebok. Each entry includes a text and a headword field. If a text has a corresponding headword, the headword field is populated; otherwise, it contains an empty string. This dataset is designed for training and evaluating headword extraction models.

Files

  • train_set.jsonl: Contains text-headword pairs from the scraped version of Nordisk Familjebok. Entries are not manually verified.
  • test_set.jsonl: Contains manually verified entries ensuring higher accuracy and quality for evaluation purposes.

Data Fields

  • text: A string representing an extraction of text (max 500 characters) from the scraped digitalized version of Nordisk Familjebok with all HTML tags removed.
  • headword: A string representing the corresponding headword. If no headword is present, this field is an empty string.

Licensing

This dataset is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0). Users can use, modify, and share this dataset for non-commercial purposes with appropriate credit and under the same license.

Citation

If you use this dataset, please cite it as follows:

@inproceedings{andersson2026atlas,
    author    = {Albin Andersson and Salam Jonasson and Fredrik Wastring and Pierre Nugues},
    title     = {ATLAS: Article Tracking, Linking, and Analysis of Swedish Encyclopedias},
    year      = {2026},
    pages     = {3566--3576},
    doi       = {10.63317/5idqekcpsxyg}
}