Transkribus and Automated Manuscript Transcription. A case study on Vittorio Alfieri’s Manuscripts
Authors
The study explores the use of the Transkribus platform for the automatic transcription of manuscripts, using texts included in the Alfieri 6 manuscript (Florence, Biblioteca Medicea Laurenziana). After an initial test phase with the generic Italian Handwriting M1 model, the results of which proved unsatisfactory, a customised model was created from the M1 model, which obtained a CER of 13.40%. A second training, expanding the ground truth set, reduced the CER to 6.9%, with more accurate transcriptions. Despite the progress, some difficulties in distinguishing between similar letters remain, but the study confirms the effectiveness of Transkribus for transcribing handwritten texts, suggesting the importance of adequate data to improve the quality of automatic transcription.
Downloads
Authors
Sara Gallegati - University of Macerata https://orcid.org/0000-0003-0553-213X
License
Copyright (c) 2025 Sara Gallegati

This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.
The Authors publishing their contributions on this journal agree to the following conditions:
- The Authors detain intellectual property rights of their work and transfer the right of first publication of the work to the journal, under the following Licence: Attribution-ShareAlike 4.0 International (CC BY-SA 4.0). This Licence allows third parties to share the work by attributing it to the Authors and clarifying that the work has been first published on this journal.
- Authors can sign other, non-exclusive licence agreements for the dissemination of the published word (e.g. to deposit it in an institutional archive or publish it in a monography), provided that they state that the work has been first published on this journal.
- Authors can disseminate their work online (e.g. in institutional repositories or on their personal websites) after its publication, to potentially enhance knowledge sharing, foster productive intellectual exchange and increase citations.
- Abstract: 134
- PDF: 44
