The State Archives Service of Ukraine has for the first time transferred data for training a national language model. These include unique historical materials, state documents, and scientific texts.
The State Archives Service of Ukraine has for the first time transferred data for training a national language model. These include unique historical materials, state documents, and scientific texts.
As the Ministry of Digital Affairs explains , in order for the national language model "Syayvo" to become a reliable source of information, it is being trained on Ukrainian data. The basis for this will be 10 terabytes of materials from the Ukrainian State Archives, which is equivalent to 70,000 books.
"We have a large array of data from different historical eras, printed and handwritten, in Ukrainian and other languages. By the end of 2026, the number of digital copies of state archives will increase from 150 million to over 200 million - this is one of the highest rates of digitization of archival heritage in the world," noted Anatoliy Khromov, head of the Ukrainian State Archives.
The process of collecting unique data for "Syavi" is currently underway. More than 50 partners, including media, universities, libraries, etc., have already provided their materials.
"As soon as the model is operational, we will publish a full list of institutions and people who helped create the national Ukrainian AI," the ministry added.
