oberbics's picture
Update README.md
b7883b8 verified
metadata
license: afl-3.0
language:
  - de
task_categories:
  - text-classification
tags:
  - history
  - migration
  - historical-newspapers
pretty_name: German Topic-specific Disambiguation Dataset Historical Newspapers
size_categories:
  - n<1K

Dataset Card for German Topic-specific Disambiguation Dataset Historical Newspapers

This dataset was specifically created for training and evaluating disambiguation methods using German historical newspapers from 1850 to 1950. It helps develop and test approaches for distinguishing between different contextual meanings of ambiguous terms in historical texts.

  • Curated by: [Sarah Oberbichler]
  • Language(s) (NLP): [German]
  • License: [afl-3.0]

Uses

  • Training disambiguation models
  • Evaluating text classification methods
  • Testing historical text processing

Dataset Structure

The dataset consists of digitized historical German newspaper articles related to remigration, with each entry containing several structured fields. Each record includes a unique document identifier that combines the newspaper name, publication date, and article number. The temporal information is provided through precise dating (date and time), and each article is linked to its digital representation via a IIIF (International Image Interoperability Framework) of the NewsEye platform: https://platform2.newseye.eu/users/sign_in

Dataset Creation

Curation Rationale

The dataset - which can be split into training and texting datesets- was compiled using the German keywords "Heimkehr" (returning home) or "Rückkehr" (returning back), which can have different meanings. The clippings were then enriched with human annotations categorizing them into relevant or not relevant.

Source Data

  • Source: Historical newspapers between 1850 and 1950
  • Primary Source: Arbeiter Zeitung, Illustrierte Kronen Zeitung, Neue Freie Presse, Innsbrucker Nachrichte
  • Primary Source Quality: OCR with varying accuracy levels and typical historical document digitization or layout recognition errors
  • Manual verification and annotation of articles and their boundaries

Who are the source data producers?

  • Austrian National Library
  • NewsEye Platform

Annotations

Human Annotation

Who are the annotators?

Sarah Oberbichler

Personal and Sensitive Information

The dataset contains only public domain historical newspaper content and has no copyright restrictions. It does contain personal information

Bias, Risks, and Limitations

  • Historical perspective and biases
  • Language and cultural context limitations
  • OCR quality variations
  • Topic-specific focus may limit generalization

Recommendations

Users should be made aware of the risks, biases and limitations of the dataset. More information needed for further recommendations.

Dataset Card Authors

Sarah Oberbichler

Dataset Card Contact

[email protected]