VEZILKA Workshop at ICT Innovations 2026

On 27 September 2026, as part of the ICT Innovations conference, a workshop organized within the VEZILKA project was held. During the workshop, the results of the work carried out so far were presented, some of which we highlight below.

The first verified benchmark for optical character recognition (OCR) of Macedonian archival documents was presented. It includes 100 pages from a total of 56 documents, with 98.8% agreement among three data annotators. An improved tokenizer for the Macedonian language was also demonstrated.

The process of creating educational content with the support of artificial intelligence and human oversight was also presented. Through this process, three free courses have been developed.

At the same time, more than 740 people have contributed through the doniraj.vezilka.ai platform. Together, they have donated more than 71,300 pages of text, over 680 documents, more than 4,000 voice recordings, and 104,000 spoken words.

Following the presentations, an open discussion was held, during which participants exchanged ideas and discussed details regarding the next steps and collaboration between the teams.