Evaluation of descriptive answers of open ended questions using NLP techniques

The COVID-19 pandemic has made a severe impact on education system. The face to face lectures attending has replaced with online learning. These closures affected the examination system as well. Answering mechanisms have become less descriptive to adapt newer modes of evaluation thus an automated system for evaluation of descriptive answers is required. This research paper introduces a mechanism for automated scoring/grading the descriptive answers for the students. It applies efficient Natural Language Processing (NLP) and Machine Learning (ML) techniques to provide a helping hand to teachers in educational sector. Three different supervised ML models are used; Support Vector Machine (SVM), Random Forest (RF) and multinomial Naïve Bayes (NB). With these, Soft Cosine similarity is being used for analyzing similarity between datasets (dataset-1 and dataset-2) and gold standard corpus. After analyzing, it is observed that Multinomial NB model outperforms on dataset-2 with 92% accuracy.