chcaa/danish-book-ads

Books and Journals in Danish Newspaper Advertisements (1800–1850)

A dataset of extracted and cleaned book titles and author names from Danish newspaper advertisements published between 1800 and 1850. The records were automatically extracted from digitized newspapers using a combination of rule-based methods, named-entity recognition, and a trained category classifier.

Dataset Description

Summary

The dataset contains 80,938 advertisement records drawn from nine Danish local newspapers. Each record corresponds to a title advertised in a book announcement: the title and author have been extracted and cleaned from the raw advertisement text, a subject category has been predicted by a TF-IDF logistic regression classifier, and the original advertisement text is included for reference.

Supported Tasks

• Text classification: predict the subject category of a book advertisement
• Information extraction / digital humanities: study the book market, publishing geography, and reading culture in early 19th-century Denmark

Languages

Danish (da). All advertisement texts are in 19th-century Danish orthography.

Dataset Structure

Data Fields

Field	Type	Description
id	string	Unique identifier in the format {newspaper}_{date}_{sequence}, e.g. berlingske_1814-12-26_1062388
date	string	Publication date (YYYY-MM-DD)
text	string	Raw advertisement text as it appears in the newspaper
title	string	Extracted book title (may still contain noise)
clean_title	string	Cleaned and normalized book title
author_original	string	Author string as extracted from the advertisement
full_name	string	Reconstructed full author name
first_name	string	Author first name
last_name	string	Author last name
category_predicted	string	Predicted subject category (see below)
confidence	string	Classifier confidence score, or "gold" (manually verified) / "author" (author-name heuristic)
category	string	Original advertisement type: book, journal, or almanac_calendar
subscription	string	Subscription or pre-order information, if present
sender	string	Sender or advertiser information, if present
nr	string	Internal sequence number within the source file
check	string	Manual annotation flag used during dataset construction

Predicted Categories

The category_predicted field uses the following labels:

Label	Description
Agriculture, household and practical arts	Agricultural, domestic, and practical manuals
Almanacs and other light readings	Almanacs, calendars
Education and Bildung	Textbooks, educational and pedagogical works
History, geography and philosophy	History, geography, travel, and philosophy
Law, politics and public affairs	Legal, governmental, and political texts
Literature and personal writings	Poetry, fiction, essays, and belles-lettres
Music	Sheet music and music theory
Religion, theology and psalms	Theology, sermons, devotional literature, and hymnals
Theatre	Theatre programmes and play scripts
Unknown	Could not be classified with sufficient confidence, neither manually

Data Splits

The dataset is released as a single train split containing all 80,938 records.

Data Collection and Processing

Source Newspapers

Records were extracted from the following digitized newspapers, accessed via the Royal Danish Library's newspaper portal (Mediestream):

Newspaper	Region
Berlingske Tidende	Copenhagen (national)
Aalborg Stiftstidende	Northern Jutland
Aarhuus Stifts-Tidende	Central Jutland
Lolland-Falsters Stifts-Tidende	Southern islands
Odense Adresse-Contoirs Efterretninger	Funen
Ribe Stifts Adresseaviser	South-western Jutland
Den Vest-Sjællandske Avis	Western Zealand
Den Nord-Cimbriske Tilskuer	Northern Jutland
Viborger Samler	Central Jutland

Citation

If you use this dataset, please cite:

@dataset{danish_book_ads_2026,
  author    = {Lassche, Alie, Rie Eriksen, Pascale Feldkamp, Katrine Baunvig and Kristoffer Nielbo},
  title     = {Books and Journals in Danish Newspaper Advertisements, 1800--1850},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/chcaa/danish-book-ads}
}

License

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Contact

For questions about the dataset, contact Alie Lassche.