Swedish Fraktur — Kraken recognition model

A Kraken recognition model for OCR of older Swedish text printed in Fraktur (blackletter). It produces a diplomatic transcription that preserves historical orthography, the long s (ſ) and period spelling.

Training data and standards

Fine-tuned from german_print (CC0) on Språkbanken's Svensk fraktur 1626-1816 (CC-BY-4.0; DOI 10.23695/5sme-7437) — 199 page images with line-level diplomatic transcriptions (double-keyed by GREPECT, digitised by Gothenburg University Library). Page images were OCR-bootstrapped and the lines aligned to the ground-truth line text (~98 % coverage) to produce PageXML for ketos train. Trained with a low learning rate (-r 0.0001); best validation accuracy 0.9880 (~1.2 % CER) on a held-out split.

Transcription standard: diplomatic — historical spelling, long-s and printed line-break hyphens are preserved; modernisation is left to downstream steps.

Attribution

Trained on Svensk fraktur 1626-1816, Språkbanken Text, University of Gothenburg, https://doi.org/10.23695/5sme-7437, licensed CC-BY-4.0.