Digitization and text recognition are key to enabling access to historical documents, including manuscripts. This article provides an overview of existing practices in the field of Handwritten Text Recognition (HTR) and several examples of foreign practices. It continues with a report on the use and training of a text model for HTR using the Transkribus tool. The National and University Library (NUK), as part of the EODOPEN project, tested the Transkribus platform on the reproduced manuscripts of seminar lectures by Janez Evangelist Krek from the early 20th century. Some of his students stenographed Krek’s lecture notes between 1903 and 1907, which were then lithographed after his review. The report describes the process of training a text model on one book and then applying the resulting model to three additional books with similar handwriting. It also presents the challenges encountered and the results, which demonstrated that the use of Transkribus is effective for recognizing texts in Slovene, although it initially requires some additional manual work. The article concludes with recommendations and ideas for further use and exploration of this technology.
|
