NewsEye Finnish Newspaper Pages: 19th Century Annotated Text

NewsEye and READ project training data comprises 200 Finnish newspaper page images from the 19th century, provided by the National Library Finland. The dataset includes carefully annotated text in PAGE XML format, produced using the Transkribus platform. It serves as a training set for historical document analysis and optical character recognition tasks.

Use Cases

• Training Optical Character Recognition (OCR) models based on annotated historical page images.
• Historical text analysis of 19th-century Finnish newspapers using the transcribed text.
• Benchmarking layout analysis algorithms using the PAGE XML format annotations.
• Studying language evolution and historical events through digitized newspaper content.

Strengths

• Contains 200 carefully annotated page images, providing a structured training set.
• Annotations follow the standardized PAGE XML format, ensuring interoperability.
• Data was produced with support from established projects (NewsEye, READ) using the Transkribus platform.
• Source material is from the authoritative National Library Finland.

Limitations

• The dataset size is limited to 200 pages, which may be insufficient for training large models.
• Metadata such as exact row count, file size, and last update date is missing from all sources.
• The provided tags from the platform (e.g., 'Meteorology', 'Geography') are inconsistent with the newspaper description, suggesting potential mislabeling.

Provenance

Source

National Library Finland (NLF), processed by Universität Innsbruck researcher Günter Muehlberger.

Collection Method

Page images were annotated using the Transkribus platform, following specific guidelines, as part of the NewsEye and READ projects.

Time Range

19th century

Geography

Finland