Evaluation of Handwritten Descriptive Responses Using Machine Learning - A Survey
Pankhuri Tayal, Parth Praveen Shetty, P Pratheeksha, R Harshita, Sreenath MV · 2023
Evaluation of student performances is crucial to the process of learning in an education system. Currently, the most common approach adopted throughout the education industry is largely reliant on human resources as forms of evaluators. As the gap in student-teacher ratios increases, this process becomes complicated and cumbersome and such manual assessment techniques lead to unintentional biases and inconsistency. Technology has been playing a role in aiding evaluators in the recent trends. Instead of the traditional pen and paper-based methods, modes of assessments are being digitized and so are the correction processes. While there are several successful technologies available for grading multiple-choice questions, there is still growing research for methods of evaluating descriptive questions which need to account for a degree of variations in student answers. For the last few decades, there have been many researchers working on developing a system that would act as a bridge between traditional methods and digitized evaluation systems, by using techniques like Optical Character Recognition (OCR). Various Artificial Intelligence and Machine Learning techniques have been proposed to evaluate descriptive answers automatically. This paper provides a comprehensive survey of such systems and analyses the drawbacks of the current studies. It has been observed that these systems do not consider essay-type questions that do not have a defined answer.