A real-time multiple-choice question generation for language testing

Abstract

An automatic generation of multiple-choice questions is one of the promising examples of educational applications of NLP techniques. A machine learning approach seems to be useful for this purpose because some of the processes can be done by classification. Using basic machine learning algorithms as Naive Bayes and K-Nearest Neighbors, we have developed a real-time system which generates questions on English grammar and vocabulary from on-line news articles. This paper describes the current version of our system and discusses some of the issues on constructing this kind of system.