Extensive Reading Automatic question generation for an extensive reading placement test more

Fri, Jul 8, 18:25-18:50 Asia/Tokyo

When practiced correctly, Extensive Reading (ER) should enable learners to become more proficient users of a foreign language. The method is only effective, however, if learners read books of an appropriate level, i.e. books which they can enjoy reading without needing to consult a dictionary. Therefore, the Extensive Reading Foundation Placement Test (ERFPT) was developed to measure the reading level of learners. The test employs comprehension questions which have been created by teachers who have volunteered their time to help develop the test. More texts and test items will improve the test; therefore, we propose a system that automatically generates questions and so reduces the burden on those making the test items. The question generation method is based on the way that the test item maker actually creates questions for the ERFPT, in other words through a process of ‘abstraction’. We use a summary created by PEGASAS, which is a transformer model for abstractive text summarization. This presentation will investigate the suitability of this method for generating questions for ERFPT. If successful this method may be used for creation of a variety of comprehension tests beyond ER.