Scientists from ITMO University, in collaboration with St. Petersburg’s Committee on Informatization and Communication and Committee on Education, created an algorithm for checking test papers that is based on a software robot by the St. Petersburg Information and Analytical Center and a neural network by ITMO’s Laboratory of Intelligent Services and Applications. 

If done manually, paper processing takes much time and effort because teachers need to input each student’s individual code, the identification number of their paper, and their score into a digital form and then upload completed forms into the Federal Information System for Assessing the Quality of Education. At a single school, teachers have to deal with an average of about 1,300 papers, which takes them several weeks to process. The neural network can do the job 20 times faster – a couple of hours instead of weeks. 

“Our smart service scans title pages of school papers via an API and then converts them into JSONs that include information on the student’s exam subject, grade, test variant, code, and score. For higher accuracy, we trained our algorithm using genuine papers and a complex method for image preprocessing. Now, teachers only need to scan title pages, launch the program, and upload the output data into the state system. Our solution, which automatically extracts and recognizes data from title pages, is the first of its kind on the Russian market,” says Dmitry Fedorov, PhD in Engineering, an associate professor at ITMO’s Faculty of Applied Informatics and the head of ITMO’s Laboratory of Intelligent Services and Applications.

Currently, the system is being put to the test at three local schools: School No. 187 and School No. 191 (Krasnogvardeysky District), and Lyceum No. 280 named after Mikhail Lermontov. 

With the help of the algorithm, teachers will be able to automate a portion of their routine tasks and focus more on their students. If successful, the technology will be introduced at other schools in St. Petersburg and the country’s different regions.