Application of Natural Language Processing - A Support Tool for Correcting Essay Questions
In the context of distance education (DE), which has been experiencing global expansion, one of the most significant challenges is the scalability of evaluating essay questions. This challenge impacts the efficiency of tutors and the quality of feedback provided to students, as grading such questions requires considerable time. To address this issue, a support system for pre-evaluation was developed, utilizing natural language processing (NLP) and machine learning (ML) to automate the assignment of preliminary grades to student responses. The proposed system is divided into two main parts - the first applies semantic similarity algorithms to compare student responses with provided answer keys; the second part is a web interface that facilitates the registration of questions and answer keys, submission of responses by students, and allows the tutor to evaluate the responses. The accuracy of the system is continuously improved through tutor feedback, enabling ML to adjust the evaluation criteria as needed. The system was tested in the course on the fundamentals of computer architecture, where it demonstrated the ability to assign meaningful grades to student responses and revealed an improvement in the accuracy of evaluations based on tutor feedback. It is concluded that the pre-evaluation support system offers a promising solution to the challenges faced in evaluating essay questions in DE environments, enhancing the scalability of the grading process and the quality of feedback provided, with potential for application across various fields of knowledge.
Abstract
In the context of distance education (DE), which has been experiencing global expansion, one of the most significant challenges is the scalability of evaluating essay questions. This challenge impacts the efficiency of tutors and the quality of feedback provided to students, as grading such questions requires considerable time. To address this issue, a support system for pre-evaluation was developed, utilizing natural language processing (NLP) and machine learning (ML) to automate the assignment of preliminary grades to student responses. The proposed system is divided into two main parts - the first applies semantic similarity algorithms to compare student responses with provided answer keys; the second part is a web interface that facilitates the registration of questions and answer keys, submission of responses by students, and allows the tutor to evaluate the responses. The accuracy of the system is continuously improved through tutor feedback, enabling ML to adjust the evaluation criteria as needed. The system was tested in the course on the fundamentals of computer architecture, where it demonstrated the ability to assign meaningful grades to student responses and revealed an improvement in the accuracy of evaluations based on tutor feedback. It is concluded that the pre-evaluation support system offers a promising solution to the challenges faced in evaluating essay questions in DE environments, enhancing the scalability of the grading process and the quality of feedback provided, with potential for application across various fields of knowledge.
BibTeX
@mastersthesis{2016-cristovao-de-lima-frinhani-aplicacao-de-processamento-de-linguagem-natural-uma-ferram,
author = {Cristóvão de Lima Frinhani},
title = {Application of Natural Language Processing - A Support Tool for Correcting Essay Questions},
year = {2016},
publisher = {Biblioteca Central da Universidade de Brasília}
}