PROMOTING ACCESS TO AFRICAN RESEARCH

Journal of Computer Science and Its Application

Log in or Register to get access to full text downloads.

Remember me or Register



DOWNLOAD FULL TEXT Open Access  DOWNLOAD FULL TEXT Subscription or Fee Access

Natural language processing techniques for automatic test questions generation using discourse connectives

Y.O. Folajimi, O.E. Omojola

Abstract


This paper presents an Automatic Question Generator from narrative text using Natural Language Processing (NLP) techniques, by paying particular emphasis on discourse connectives. Discourse connectives are words or phrases that indicate relationships between two logical sentences or phrases and suggest the presence of mutually related extended verbal expression. A detailed design and development issues are discussed in this work. The system formulates questions from a  student‟s lecture material and displays it to the user in the form of test questions. It also makes provision for the user to answer the questions in essay format and on submission of the answer, the system grades the user and returns the score  obtained to the user. The questions were generated by first extracting the text from the materials supplied by the user using text processing concept of NLP. The  discourse connectives in each of the sentences are identified and questions are generated from the sentence based on the peculiarity of the connective contained in the sentence. The marker checks the submitted answer to see if it contains 50% of the expected answer. If it does, the user‟s answer against a particular question is marked as being correct, otherwise it is marked wrong. The end result is an Automatic system which generates test questions for the user and allows him/her to submit essay answers back into the application system. Evaluation results with the system show that the generated questions achieved average accuracies of 87.5% and 88.1% by two human experts.

Keywords: Discourse Connectives, Machine Learning, Automatic Test Generation E-Learning.




AJOL African Journals Online