Nada Ahmed, Mariam Yasser, Rawan Ahmed and Philip George
5 January 2022
The grading of short answer questions has become a headache recently. As a result, the system’s main goal is to build an automatic correction model for subjective questions, which will make the process of checking answer essay questions easier. The input answers are reference answers and the student answers then the system will correct and give the score. The system has been categorized into some steps firstly, preprocessing of data using tokenization, stop word removal, porter stemmer, Then, using TF-IDF, a frequency calculating approach, these answers were transformed into vectors. Then we utilised Cosine similarity to determine how similar the answers were. To assess the proposed system’s performance. To improve accuracy, we applied latent semantic analysis (LSA).
1.1 Purpose of this document
The purpose of the document is to define the requirements of the application and to set guiding rules to help through developing the application. The document’s goal is to collect and analyze all of the various concepts that have surfaced to define the system. We aim to make correcting exams and papers easier for both instructors and students.
1.2 Scope of this document
This document provides out the requirements and the system boundaries that must be met. It explains the classes and their relationships, as well as the database, dataset, and its structure, as well as the comprehensive functional and non-functional requirements also the main functionalities of our system.
1.3 System Overview
Our system is designed to correct subjective and essay questions automatically. It acts as a co-assistant to the professor while correcting these types of questions. First, the professor shall set up the essay questions for assignments or exams. Then, the students shall reply to the assignment. The system then shall analyze and scan the students’ answers and proceed to evaluate them according to some keywords. The system shall analyze the text for possible errors and correct it automatically. We began by preprocessing the text, which included stemming, removing stop words, lower case, and tokenization. We next utilized TFIDF in feature extraction , and as an initial step, we compared the student answer to the reference answer using cosine similarity. but cosine similarity is compared word by word so we added LSA Latent Semantic Analysis to understand the context and compare the student answer with the reference answer. The system shall use various machine learning techniques to achieve its goal.
1.4 System Scope
1-For subjective questions, auto-correction is seen as a solution to tackle one of academia’s most known problems.
2-Professors used to spend a lot of time on their own correcting essays and subjective questions.
3-The aim is to develop a system that uses a certain technique to match synonyms and keywords, then checks the student’s answer to determine if it’s correct.