Image
depiction of the siege of Smolensk
The siege of Smolensk 1609-1611
Breadcrumb

The Window on History: AI-Assisted Deciphering of the Smolensk Archive

Research
Culture and languages

Welcome to a seminar on a pilot project exploring how AI-based handwritten text recognition can help make older Russian archival materials in Swedish collections more accessible.

Seminar
Date
11 Jun 2026
Time
15:15 - 17:00
Location
Room C442, Humanisten, Renströmsgatan 6 and Zoom

Participants
Thomas Rosén and Erika Rydergård
Good to know
Language: English

For Zoom link, please contact antoaneta.granberg@sprak.gu.se
Organizer
The Department of Languages and Literatures, the research area Language in Society, and the Slavic Research Seminar.

Abstract

At this seminar, Erika Rydergård and Thomas Rosén will be presenting their experiences and results of a pilot project carried out during the first half of 2026. The Window on History: a Pilot Project to Promote the Availability of Older Russian Materials in Swedish Archives (Fönstret till historien: ett pilotprojekt för att tillgängliggöra äldre ryskt material i svenska arkiv). The project was funded by the The Royal Swedish Academy of Letters, History and Antiquities.

The main aim of the project has been to create a Handwritten Text Recognition (HTR) model capable of transcribing early seventeenth-century Russian cursive writing (skoropis). This work has largely been carried out on the Transkribus platform, which uses AI to transcribe historical documents.

The documents that were processed within the project, stem from the Russian city of Smolensk, where they were held at the Governor’s office. During the years 1609–1611, Smolensk was besieged by Polish-Lithuanian troops. The material, which represents a rich variety of administrative documents as well as private letters, ended up in Sweden later in the seventeenth century and has remained there ever since. Today, the Smolensk papers form part of the Skokloster collection of the Swedish State Archives and have been fully digitised. The HTR model also incorporates data from the Novgorod Occupation Archive, transcribed with support from the National Archive of Sweden.