The project aimed to make Abraham Kuyper’s many writings, predominantly correspondence, accessible to researchers. The team also investigated how AI can be used to automatically convert scanned letters into readable files. This approach is particularly relevant given that the Historical Documentation Centre (HDC) at VU Amsterdam manages approximately 100 boxes of Kuyper archive material while interest in his work remains strong. Additionally, about 4 kilometers of other archival material at the HDC awaits digital disclosure.

The project utilized Transkribus, a widely used European AI tool within archival institutions. The model was first fed with manual transcriptions, with documents being deciphered word by word. Bleeker describes this as painstaking work, especially with handwriting as difficult as Kuyper’s. Subsequently, scanned documents were automatically transcribed and checked for errors. That feedback was then returned to the model to refine it further.

The project is now nearing completion. The result is an AI model that can transcribe Kuyper’s handwriting with less than 10% error margin, which is considered an excellent score. The model can also be used outside the VU archive, making Kuyper’s legacy accessible elsewhere in the Netherlands regardless of where documents are located. The knowledge and experience gained also make it possible to unlock other archives faster and more effectively in the future, such as archives containing correspondence from administrators and professors affiliated with VU Amsterdam.