Improving oral reading fluency assessment using automatic speech processing technologies

Using speech processing technologies with third-grade readers, several new measures of reading rate were developed that involved measuring reading speed of individual words within the context of continuous text. Scores from one test form consisting of three passages were correlated with scores from another test form with three different passages to evaluate each measure’s reliability. Correlations were also compared with reliability of median Words Correct Per Minute (WCPM) across the two test forms. The findings indicate that one of the new measures, NM5, is more stable across passages and more reliable than WCPM. Another experiment with fourth-graders replicated the result, suggesting that machine-generated scores can offer more precision and better reliability than WCPM, the commonly used measure of oral reading fluency. Implications for oral reading assessment are discussed.