This function appeared in the Yandex service “Search archives”. The neural network was taught to recognize handwriting features, recognize letters that lost their relevance, and understand the special structure of archival documents. If you hover over the desired part of the document, it will be immediately highlighted in the digital copy. The first archive added to the service was the Moscow Main Archive – it was on materials that the developers trained the neural network.
Currently, the archives of the Orenburg and Novgorod regions have been added to the database. The amount of data will increase over time.
Documents available for recognition include church records, confessional documents, revision stories containing the results of the 18th to early 20th century census.
Source: Ferra