An intelligent system for exam generation

Abstract

This paper describes an intelligent system for exam generation. The supposed system consists of two main
subsystems, namely Question Generation 'QG' and Exam Maker ‘EM’, which are designed to produce an exam
sheet for an E-course based on modules in higher education institutions.
The first subsystem, QG depends on natural language processing techniques such as tokenization, part of speech tagging (POS) and Named Entity Recognition (NER). It also uses the template matching approach to generate two types of questions: factual questions such as WH questions (who – where – when – what and what is the percentage of), and gap filling question GFQ.
Exam maker is the second subsystem in the proposed system, and it generates an exam sheet taking into account a set of criteria which includes the relative weight of each module, the objectives set for each lesson, formal quality assurance standards of the exam paper, and diversity of the questions.
The proposed system has been implemented in the Pharmacognosy E-course taught to the first year students in the Faculty of Pharmacy, Mansoura University, Egypt. Evaluation of the proposed system is presented.
Key words: Natural Language Processing, Question Generation, Tokenization, Part Of Speech Tagging, Named Entity Recognition.