Kat -57 ground truth dataset

Background

Catalogue -1957 is an alphabetic library catalogue listing Lund University Library's holdings up to 1957. A project has been ongoing to scan and transcribe the catalogue cards.

About 10.000 cards were manually transcribed to create a ground truth dataset.

From 2178 card drawers, one drawer for every letter in the alphabet was selected to transcribe, except for in the letter S, where two drawers were selected.

The writing on the catalogue cards is a mix of typewriter and handwriting. There are more than 10 different hands.

The cards were transcribed by a small team at the University Library.

Dataset

The set consists of PNG images with corresponding PAGE XML files. The transcriptions were made in eScriptorium.

File structure description

• images/: PNG images
• pagexml/: Corresponding Page XML files
• README.md: Background and information