Deep Learning in Automation of Checking Homework Assignments

Abstract

The article deals with the current issues of knowledge management automation. The authors describe the possibilities of deep learning methods for checking homework and control tasks in such school subjects as Russian language, literature, social studies, history. As a rule, the main form of presenting answers to tasks is text. In this regard, it is proposed to automate the process of recognizing handwritten students texts and checking for compliance of the student’s response with the template document proposed by the teacher for this task. The main process of the service is handwritten text recognition technology, which is implemented on the basis of convolutional and recurrent neural network architecture using a decoding algorithm based on connective time classification. The article also provides a description of the online service, which implements the ability to download the answer to the task by students, recognition of the answer, determination and output of the answer similarity percentage with the attached answer from the teacher. According to the authors, this service will automate the teachers routine operations to check homework. It will be especially useful during the implementation of distance learning during pandemics, quarantines, etc.