Development of a readability index attuned to the new English Course of Study in Japan

Ozasa, Toshiaki and Weir, George R.S. and Fukui, Masayasu (2014) Development of a readability index attuned to the new English Course of Study in Japan. In: Proceedings of ICERI2014 Conference. 7th International Conference of Education, Research and Innovation . International Academy of Technology, Education and Development (IATED), pp. 2446-2453. ISBN 9788461724840

[img]
Preview
PDF (Ozasa-etal-ICERI2014-development-of-a-readability-index)
Ozasa_etal_ICERI2014_development_of_a_readability_index.pdf
Accepted Author Manuscript

Download (289kB)| Preview

    Abstract

    For several years, we have been engaged in evaluating techniques for estimating readability. Arising from this work, there have been several iterations of readability indices attuned to the English textbooks employed in Japanese high schools for the English Course of Study. In this paper, we briefly review approaches to readability measurement before detailing our project that aims to develop new readability indices that are attuned to the new English teaching system of Japan. Our earlier approach to the Japanese English teaching context employed regression analysis to develop a linear measure in which sentence length, word length, word difficulty and idiom difficulty are independent variables and the textbook year of passages is a dependent variable. This linear function (Diff) proved to be broadly effective as a guide to reading level for English texts in the Japanese school context and was released for use as 'the Ozasa-Fukui Year Level'. From 2012, changes to the English Course of Study in Japan made it necessary to update this English readability index. To this end, our earlier approach was enhanced to provide a non-linear formula in which an experts’ readability judgement of English sentences is employed as a dependent variable and combined with the previously developed linear function (Diff) as an independent variable. The resultant computation yielded a non-linear function expression: NewDiff = a/(1+b*exp(-c*Diff))+0.9 in which we have the following component values: a=3.8593, b=766.9372, c=2.5709, Diff = 0.0863*Words/S + 0.2943*Syllables/W + 0.6332*WordDiff/W + 0.0665*IdiomDiff/S + 0.5366. Applying this new measure, we determined that the coefficient of determination for this new formula (r^2) was 0.8236. Thereby, indicating a high degree of fit to the target data.