Full Title: ‘Does Handwriting Text Recognition Work for Damaged Archives?'
By Marco Roling, 2020, Indepedent Research
Abstract
Handwriting Text Recognition (HTR) is used on a large scale for digitized archives, but so far experiments have focused on manuscripts with a high standard of preservation and legibility. This paper describes some controlled experiments done on text samples with various types and degrees of archival damage, in order to assess their suitability for HTR.
Also some ideas are expressed about how to predict the success of HTR when it is applied to large volumes of scans. Lastly, it is suggested to enhance scans before subjecting them to the HTR process, with the intention to further improve the overall quality of automated transcriptions.
Read and download the full version in English (PDF) >>