Assessing the value of one's answers constitutes one of the most essential aspects of learning. In modern day and age of digital technology, there have been produced a great deal of systems that are able to automatically handle evaluation. Long answers and brief answers are the two primary varieties of this type of response. Some of the existing scoring systems that are used on lengthy answers have demonstrated outcomes that are around average when it comes to putting a score on the student's answer. The information retrieval approach is employed in such systems to determine the degree of similarity between the answers provided by the students and the answers provided by the references; nevertheless, such scoring methods do not yet produce the best possible results. Each answer in a short answer just contains a select few keywords. The evaluation of such brief responses, which only contain a small number of keywords, calls for a unique approach, particularly when it comes to the weighing procedure. Cosine, dice, Euclidian distance, and overlap are the four metrics that have been analysed in this particular piece of research. To begin, the texts need to be transformed into some form of numerical expression. In order to accomplish this goal, the text was first broken up into n-grams at the word level, and then a bag containing all of these n-grams was constructed. After the frequencies of the n-grams have been computed, the frequency matrix of the dataset can be constructed. In order to produce a bag of n-grams, numerous filters, such as stemming algorithms, stop word removers, number and comma removers, etc., are utilised. The entire experimental inquiry was carried out with a dataset consisting of plagiarised long replies from a corpus

Automated Essay Evaluation, Automatic Scoring, Automatic Grading, String Similarity, Content-Based Similarity, Natural Language Processing