Comparing Passage Length and Human Versus Speech Recognition Scoring of Oral Reading Fluency

• In Event: Poster Session 11
  In Poster Session: Division H Poster Session: Assessment in Schools

Abstract

The purpose of this study is to compare traditional oral reading fluency (ORF) measures to a computerized oral reading evaluation (CORE) system that uses speech recognition software. We applied a mixed model approach with two within-subject variables to test the mean wcpm score differences and the error rates between: passage length (25, 50, 85, and 250 words), and scoring method (real-time, recorded audio, and automated speech recognition [ASR]). We also used weighted Cohen’s kappa to analyze the word score agreement across scoring methods. We found that shorter passages had greater means than traditional ORF passages, and that real-time (traditional) scores had greater means than both audio recording and ASR scores; the latter two were similar across passage length and grade.

Authors

• Joseph F.T. Nese, University of Oregon

• Akihito Kamata, Southern Methodist University

• Julie Alonzo, University of Oregon