• News
  • Society
  • Ukraine training AI to recognize doctors' handwriting
1686

Ukraine training AI to recognize doctors' handwriting

Artificial Intelligence. Collage: Ukrainian News
Artificial Intelligence. Collage: Ukrainian News

The Ministry of Economy of Ukraine, together with the State Archives and other departments, has begun creating a national dataset of Ukrainian-language handwritten materials for training artificial intelligence (AI) systems. The goal is to teach the language model to parse old manuscripts, including medical certificates.

The details of the project were told by ML Lead of the "Mriya" application and artificial intelligence advisor at the Ministry of Economy Dmytro Voytekh in the AI&I podcast.

The implementation of this initiative should accelerate the digitalization of public services and promote the digitization of archival materials.

The need to create such a dataset arose during work on the "eDozvil" project, which involves transferring the process of issuing licenses for entrepreneurs to a digital format through "Diia". To do this, the algorithms need to analyze the documents submitted by applicants.

ADVERTISING

In particular, to obtain some licenses, you need to upload diplomas issued in the 1990s, which are often filled out by hand, have low image quality or are damaged. As Voytekh noted, existing OCR systems do not provide proper recognition of such materials in Ukrainian.

According to the developers, there are no high-quality marked-up corpora of Ukrainian handwritten text in the open access, which are necessary for training models. Therefore, the Ministry of Economy initiated cooperation with other state institutions to quickly form an appropriate database.

The key partner was the State Archives of Ukraine, which is interested in using technology to digitize millions of pages of historical documents and simplify access to information.

Meanwhile, about 10,000 writers have published a book "Don't steal this book", which contains only a list of their names, to speak out against the possible use of their works by artificial intelligence companies without permission. The publication was distributed to visitors to the London Book Fair on March 10.

ADVERTISING

Who we are: About us, Contacts. How we write news and our principles: Editorial code. We did our best. If you found this valuable – please support us.

To request a correction, please send an email.