Performing Semantic Analysis of Short Descriptive Answers using NLP

The concept of semantic similarity has garnered significant attention within the domains of artificial intelligence and semantic webs. The quantification of similarity in relation to words, brief textual passages, paraphrasing, text summarization, and automated grading has garnered significant attention in contemporary times. This paper outlines a methodology for assessing student responses by comparing them to a teacher's exemplar response, and subsequently evaluating and validating machine-generated outputs against human assessments. The brief passage must align semantically with a modelled response in order to receive a corresponding grade. The present paper presents semantic measures that rely on the similarity between individual words, with the relatedness of two words being measured through the use of Word Net taxonomy. Also, the author aims to provide a comprehensive evaluation of the proposed system by implementing it on answers dataset.