Paper Key : IRJ************895
Author: Uday Popat Dube,Mayur Bhimraj Dhumal,Mayank Sanjay Hangshoo
Date Published: 08 Nov 2024
Abstract
Subjective Answer Assessment aims to address the complexities of evaluating written content, which is often time-consuming and can lead to inconsistent and unfair scores. We propose an innovative method that uses machine learning and linguistic techniques to automate this assessment process. We specifically leverage tools such as WordNet, Word2Vec, Word Movement Distance (WMD), Cosine Similarity, Multinomial Naive Bayes (MNB), and Term Frequency-Inverse Document Frequency (TF IDF) to assess the quality of responses. Our approach combines problem-solving concepts and core concepts to assess student responses, allowing machine learning models to predict scores based on content and response quality. Initial results show that WMD performs similarly accurately over cosine, with up to 88% error reduction without the MNB model and an additional 1.3% error reduction when MNB is included. The project aims not only to demonstrate the potential of electronic systems in educational assessment, but also to improve the overall effectiveness and integrity of contextual assessment.