The Ministry of Economy of Ukraine, together with the State Archives and other agencies, has initiated a large-scale project to collect a database of Ukrainian-language handwritten data. The main goal is to train artificial intelligence to qualitatively recognize complex handwritten text, old documents, and medical certificates.
The Ministry of Economy of Ukraine, together with the State Archives and other agencies, has initiated a large-scale project to collect a database of Ukrainian-language handwritten data. The main goal is to train artificial intelligence to qualitatively recognize complex handwritten text, old documents, and medical certificates.
The details of this initiative were shared by Dmytro Voytech, ML Lead of the Mriya application and AI advisor at the Ministry of Economy, in the AI&I podcast. According to Voytech, the initiative will significantly accelerate the digitalization of public services and open the way to global digitization of historical archives.
The idea of creating a national dataset was born while working on the e-Permit project, which aims to digitize the issuance of licenses for entrepreneurs through Diya. To automate this process, algorithms need to analyze applicants' documents.
However, it turned out that to obtain many licenses, it is necessary to upload old diplomas (some dating back to the 90s), which are often filled out by hand, poorly photographed, or have defects. According to Dmytro Voytech, ready-made OCR solutions (optical character recognition systems) that exist on the market turned out to be completely powerless against Ukrainian manuscripts.
"We encountered the fact that it works very poorly on Ukrainian manuscripts, especially considering that our first licenses are related to medical services. We all understand what the font of our beloved doctors looks like," said Vojtech about the problems of Ukrainian handwritten texts.
Faced with this problem, the developers realized that there simply were no high-quality and marked-up corpora of Ukrainian handwritten text in the open access. In order not to wait years for the eDozvil system to independently accumulate a sufficient amount of data, the Ministry of Economy used its authority to join forces with other government agencies.
The largest partner of the initiative is the State Archives of Ukraine. This institution has a huge interest in the development of technology, because their strategic goal is to digitize millions of pages of historical documents. Instead of spending hours searching for information physically, as is the case now, a high-quality AI model will allow archives to be transformed into a convenient knowledge base, where information can be searched for as easily as in a search engine.
