Automatic Multiple Choice Question Generator by Using Machine Learning

Published in Institut Teknologi Bandung, 2017

Type of Paper

Academic Paper for lecture of IF3280 Socio Informatics and Professionalism
Computer Science, Institut Teknologi Bandung
Bandung, Indonesia


English reading comprehension is one of the most required skill for people whose native language is not English. One example is when you study in an English spoken country, you have to read many literatures supporting the course in English language. Therefore, your ability to get the concept behind the literatures effectively determines your level of understanding towards the literatures. To achieve this, many universities or companies require the TOEFL, IELTS, or TOEIC score in order to know your basic English skill. Because of this, the applicants of university or company decide to take one of those exams and do many preparations by studying from any resources providing the practice questions, such as the English exam books or resources from the internet. Unfortunately, those resources provide the limited number of sample questions and the genre of article cannot be customized by the learners. This makes the learning process become not fun and not effective at once since the learners cannot set their preference in the literature’s genre and this makes them cannot improve their understanding in the other genres. This paper proposes an automatic multiple-choice question generator for English document. It runs on LINE platform as a bot so that the learners can interact directly and learn in more effective way. To achieve that, it uses LINE Messaging API to receive the user’s query and send the reply back to the user. The system does several things to generate the best questions, such as sets the blank position for the answer to every word in the document, determines the most reliable sentences (which has already had the blank position for the answer) to become the question, and generates the four options as the possible answers.

Download paper here