www.ijcrt.

org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882

AUTOMATIC MCQ GENERATION

Prajakta, Vivek, Pragati, Yogesh and Geeta Atkar
Department of Computer Engineering Department of Computer Engineering
Student at GHRCEM, Wagholi. Professor at GHRCEM, Wagholi
Pune, Maharashtra India,412207 Pune, Maharashtra, India 412207

Abstract learning process. These procedures must be

We propose classification of abstracts by
compatible with other solutions that provide
considering their context. In proposed approach,
we initially classify the documents. We find the personalized feedback to students for understanding
context of an abstract by looking for associated and improving the quality of their learning experience.
terms which help us understand the focus of the Many e-learning proposals use Multiple-Choice
abstract and interpret the information beyond
Questions (MCQ) as an assessment tool. Basically, an
simple keywords. The results indicate that context
based classification increases accuracy of MCQ consists of a question text and a few (e.g. four)
classification to great extent and in turn discovers choices, from which one is the correct answer and the
different contexts of the documents. Further this
others are incorrect alternatives (called distractors).
approach can found to be very useful for
applications beyond abstract classification where MCQs are labor intensive, time consuming and
word speaks very little and lead to ambiguous state difficult to construct. For this reason, recent efforts
but context can lead you to right
have focused on the automatic generation of well-
decision/classification. Automatic question
generation is part of NLP i.e. Natural Language constructed MCQs, mainly for vocabulary assessment
processing, is an area of research where many or grammar exercises. The automated creation of tests
researcher have presented their work and is still an involves generating distractors based on certain
area under research to achieve higher accuracy.
Many researcher have worked in the area of knowledge and, subsequently, using these distractors
automatic question generation through NLP and to create the assessment test.
numerous techniques and models have been
developed to generate the different types of 2.RELATED WORK
question automatically and in many different Different dataset is used by different researchers
languages work has been done like English, to analysis the result through sentiment analysis.
Punjabi, China, Spain, etc. Some work is discussed in this section.
.
1. A Taxonomy of Questions for Question Generation.
Key-words: Data analytics, Sentiment Analysis,
Center for Computational Language and Education
News Articles ,LSTM-RNN,Arima, Deep-
Learning. Research, University of Colorado, Boulder
Rodney.Nielsen @ Colorado.edu. Rodney D.
Nielsen1, Jason Buckingham, Gary Knoll, Ben
1. INTRODUCTION Marsh and Leysia Palen March 2014. - In
New e-learning methodologies require summative assessment, questions are intended to
assessment procedures that automatically measure the evaluate the answerer’s knowledge, understanding
students’ achievements during the teaching and and skills.

IJCRTI020031 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 152

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882
2. Optimizing the Correction of MCQ test answer time consuming for both – the writing and
sheets using Digital Image Processing. 2016 Eighth assessment. To overcome this, we have Multiple-
International Conference on Information and Choice Examinations (MCQs) which is a set of
Knowledge Technology (IKT), Hamedan, Iran. questions with a number of alternatives one of
978-1-5090-4335-4/16/$31.00 ©2016 IEEE- Image which is the correct answer. Hard maps in order to
segmentation of our algorithm can be improved by determine the difficulty level of a question.
choosing larger dimensions for neighboring Existing approaches to MCQ generation primarily
matrices to resolve the recognition error occurring focused on generating questions from a single
on the answer sheet. Also the software could be sentence. However, the text might contain certain
ported to smartphones that will allow teachers to facts that are expressed through multiple sentences.
travel to test centers carrying only printed tests with While studying human-generated MCQs in
no need for additional resources, apply the tests, different domains, we found that many MCQs deal
capturing images and calculating scores in locoing with such multi-line facts.
and being able to discuss the test results with the
students on the same day.
3. Neural Models for Key Phrase Detection and
Question Generation. 1MILA, Universit´e de
Montr´eal 2Microsoft Maluuba February 2018.
Sandeep Subramanian1,*,y, Tong Wang2, Xingdi
Yuan2, Saizheng Zhang1,y, Adam Trischler2, and
Yoshua Bengio1.- Our model first estimates the
probability that word sequences in document
compose “interesting” answers using a neural
model trained on a question answering Or
extractive nature of answers in documents.
4. METHODOLOGY
4. Literature review of automatic question generation
systems. School of Science, RK University
Automatic Question Generation Systems takes
**Associate Professor, College of Agricultural
a natural language text as input and generate
Information Technology. Sheetal Rakangor*, Dr.
questions of mcq types and scopes for the
Y. R. Ghodasara** January 2015 -proposed
user. In particular, we focus on the problem of
framework helps in question generation by
automatically generating factual questions
deploying agents, the agents will perform various
from individual texts. Serious management
operations like document processing, information
problems concerning assessment using on-
classification and question generation.
paper tests, especially to courses with a high
3.Proposed model: number of enrolled students (ranging from 200
to 800).
The various modes in which exams are
conducted have evolved over the period of time.
The traditional method of evaluating student’s
performance was based on
Descriptive/Constructive Response type of test.
The major disadvantage of such exam is that it is

IJCRTI020031 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 153

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882
5. MODELING AND ANALYSIS

• User/Student
1. Registration
2. Login
3. Give MCQ Exam
4. View Result
5. View Progress

• Teacher
1. Registration
2. Login
3. Upload PDF or Text file
4. Select word for MCQ answer.
5. Generating MCQ
6. Select MCQ from generated MCQ
Dataset.
7. View student progress
8. Suggest study material to student
according to student progress.

 Admin

1. Login
2. Maintain Exam Details
3. View Teachers
4. View student progress
5. RESULTS AND DISCUSSION

6.FUTURE WORK:
The scope of the generated questions is either
specific or medium. Automatic Question
Generation Systems that take natural language text
as input and generate questions of various types
and scope for the user. The best thing about this
system is that make processing of generating
Objective Type paper easy way and analyzing data
generated by MCQ tests in order to overcome the
existing problems of such test and thus help in
enhancing the student’s performance.

IJCRTI020031 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 154

www.ijcrt.org © 2021 IJCRT | Volume 9, Issue 11 November 2021 | ISSN: 2320-2882
of Colorado, Boulder
7.CONCLUSION [email protected] March 2014.
[2] Optimizing the Correction of MCQ test answer
Multiple choice questions of fill in the blank type and
sheets using Digital Image Processing
analogous type were successfully generated. Questions
[3] Neural Models for Key Phrase Detection and
of good standard were produced with reasonably high
Question Generation. 1MILA, Universit´e de
accuracy (higher compared with the existing
Montr´eal 2Microsoft Maluuba February
models).Evaluation is essential in the teaching-
2018.
learning process and MCQs are popular for
[4] Literature review of automatic question
educational assessment.
generation systems .School of Science, RK
We established of six broadly classified dependent
University **Associate Professor, College of
phases, namely, pre-processing, sentence selection,
Agricultural Information Technology. January
key selection, question formation, distractor
2015
generation, and post processing.
[5] Ontology-Based Multiple Choice Question
Generation.Received: 15 May 2015 /
8.REFERENCES
Accepted: 5 October 2015 / Published online:
17 November 2015
[1] A Taxonomy of Questions for Question [6] Nlp module for bulgarian text processing
Generation. Center for Computational .international conference on innovations in
Language and Education Research, University science and education 22-24, 2017.

IJCRTI020031 International Journal of Creative Research Thoughts (IJCRT) www.ijcrt.org 155