Neural Network Grading: Automated Evaluation of Theoretical, Mathematical, and Diagrammatic Responses
Yogita Hande, Ritika Pandey · 2023
Neural network-based computerised paper evaluation has the potential to completely transform the current paper evaluation procedure. The system may grade student papers automatically with the aid of machine learning algorithms, which can decrease human mistake and boost efficiency and speed in the grading process. The ability of neural networks to handle a variety of questions, including both subjective and objective questions, is one of the main benefits of utilising neural networks for computerised paper evaluation. The approach outlined in this paper utilizes Optical Character Recognition (OCR) to process images and extract various components. Various techniques are used to prepare the OCR for specific types of questions. Theoretical questions are processed using the Artificial Neural Network algorithm, while mathematical expressions and diagrams are evaluated using the Convolutional Neural Network. In order to achieve this, the CNN must recognize specific characteristics in the diagrams and mathematical expressions and assign scores based on the accuracy of the solution provided. The neural network is trained to recognise patterns and features in the data using the backpropagation technique. This method alters the weights as well as the biases in the neural network to lessen the difference between the expected output and the actual result. Until the network's performance improves and it can precisely assess the student's performance, this process is repeated across a number of epochs. This strategy is especially successful since it enables the system to examine a vast amount of data and anticipate future events based on patterns in that data.