コンピュータソフトの音読評価その妥当性と評価特性の検証 : アメリカ人評価者の評価との関係

概要

論文の詳細を見る
Some studies have tried to assess the learners' voice data of reading aloud with computer software. However, few referred the validity of the scores which computer software produced. The study examined the validity and features of the scores of computer software. Four American raters (four males) assessed the voice data of 32 Japanese university students in measures of word recognition, stress, pause, and intonation. The authors compared the score of the software with the measures of raters in addition to the rate of delivery of the voice data and learners' proficiency. The results of t-test showed that the score was correlated relatively high with all the measures of the assessments except word recognition (study 1). In study 2, the assessments and scores in study 1 were reexamined. The results of structural equation modeling indicated that the models did not show good fit indexes when the score was incorporated. The authors discussed that software needed to have the different role from teachers because of the differences of the features between raters' assessment and the scores of software.