Berlin State Library OCR Texts from 1470 to 1945

Aggregating 4,988,099 full-text pages from 28,909 OCR-processed historical works. The texts originate from the digital collections of the Berlin State Library, covering a time period from 1470 to 1945. The language of each page has been automatically identified.

Use Cases

• Train or fine-tune multilingual language models on the identified language labels and historical text sequences.
• Analyze linguistic evolution and vocabulary usage across the 1470-1945 time range covered by the texts.
• Develop OCR post-correction models using the machine-generated text from the 28,909 processed works.
• Conduct historical research by analyzing thematic content within the 4,988,099 pages of digitized material.

Strengths

• Large scale of 4,988,099 full-text pages.
• Covers a long historical period from 1470 to 1945.
• Multilingual content with machine-identified language labels.
• Derived from 28,909 distinct, digitized historical works.

Limitations

• OCR text may contain recognition errors common in historical documents.
• The dataset represents only a subset (28,909 of 153,942) of the library's digitized collection.
• Language identification is automated and may have inaccuracies.

Provenance

Source

Berlin State Library (SBB) digital collections.

Collection Method

OCR processing and language identification (langid) applied to digitized historical works.

Time Range

1470 to 1945.

Freshness

Last updated in August 2022.

Geography

null