The models were trained upon roughly 2500 pages (ca. 400k words) of handwritten transcriptions from the School of Scottish Studies Archives (https://www.ed.ac.uk/information-services/library-museum-gallery/cultural-heritage-collections/school-scottish-studies-archives). The transcriptions are from recordings of hero tales and international tales in Scottish Gaelic, made between 1949 and 1979. They were gathered from tradition bearers living in the Scottish Highlands and Islands. The majority of the transcriptions are from one hand (~80%) with the remainder split across approximately 10 hands. Thus, the generalisability of the models is limited at present. We hope to diversify the training data in the future by adding further hands.
- TranskribusTranskribus
Everything about Transkribus
Download TranskribusDownload the Expert Client
Transkribus LiteUse Transkribus in your Browser
Credits & PricingBuy Credits for Handwritten Text Recognition
Transkribus.aiText recognition with the click of a button
How-To GuidesLearn how to use all Features
Public AI ModelsExplore all publicly available models
Scholarship ProgrammeSupporting future scholars
metagrapho apiHandwritten Text Recognition api for Transkribus
Transkribus User Conference 22Connect with like-minded people
- ScanTent
- read&search
- The COOPOur Members
See all of our great Members
Our TeamMeet the READ-COOP Team
Success StoriesHave a look at some exceptional Projects
- Resources