The Utility of Corpora for Composing Reading Comprehension Questions for Large-Scale Tests

Abstract

This paper examines the utility of corpora in the composition of questions for large-scale tests. After explaining the present situation of corpus use in the field of language testing, I explain some of the advantages of using corpora in the process of composing questions. Next, in the context of reading comprehension questions, I perform a discriminant analysis of corpus data according the levels of the Japanese Language Proficiency Test, using the reading comprehension texts from the Japanese Language Proficiency Test as learning data, and data from the Balanced Corpus of Contemporary Japanese as evaluation data. Based on this analysis I identify the levels to which the Balanced Corpus data corresponds, and the extent to which such data can be used in the composition of test questions. In conclusion, I also indicate the importance of taking a qualitative approach to analysis that examines concrete examples, in addition to the quantitative approach described above.