Purpose: The team fully developed and tested Moby.Read, a tablet-based oral reading fluency assessment. Reading out loud is a foundational skill for students in grades 1 to 5 and is included in state learning standards. In early grades, oral reading fluency assessments are often used to place students, monitor progress, and guide teaching. However, traditional oral reading fluency assessment is time-intensive for teachers, and the assessments typically include several sources of uncontrolled variance reducing validity and reliability. Project Activities: Through a 2014 contract from the NAEP program at the National Center for Education Statistics and with SBIR Phase I funding (completed in 2016), the team developed a prototype of Moby.Read, a tablet-based app where grade school students read a passage aloud and the speech-recognition software generates an assessment of oral-reading fluency in real-time. In a pilot study at the end of Phase I, the researchers tested the prototype with 99 grade school students and four teachers. Results showed Moby.Read generated scores correlated with traditional human-rated oral reading fluency assessments, and that most students indicated preferring Moby.Read to teacher-administered tests. Phase II focused on extending content, including Spanish text, adding more diagnostic scoring, and implementing cloud-based tools for aggregation, review, and management of scored performance samples. After development was completed, researchers conducted a pilot study with 130 grade 4 students to assess the feasibility and usability, fidelity of implementation, and the accuracy of Moby.Read for assessing reading fluency compared to traditional human-scored assessments. Product: Moby.Read is a tablet-based speech recognition app that grade school students use to administer oral-reading fluency assessments in real-time. The app is intended to replace face-to-face oral reading assessments done by teachers, and to save time and increase the accuracy of the assessment. Through the teacher dashboard, Moby.Read aggregates assessment results at the class and individual student level, and provides teachers detailed reports on types and frequency of errors to inform instruction. Moby.Read is implemented in and out of classrooms to supplement oral reading instruction and improved fluency. Video Demonstrations: Moby.Read Phase I Prototype: https://youtu.be/4gb0nmbVfQo SkillCheck Phase I Prototype: https://youtu.be/_97B3FUAA9Q Recognition: 2020: Finalist for an SIIA Codie Award Related IES Projects: Automated Basic Reading Assessment (EDIES16C0004); Early Reading Diagnostic Profiler (91990020C0070 ); SkillCheck: Automatic Early Reading Diagnosis and Intervention (91990021C0040) PUBLICATIONS Bernstein, J., Cheng, J., Balogh, J., & Downey, R. (2020). Artificial intelligence for scoring oral reading fluency. In H. Jiao & R. Lissitz (Eds.), Applications of artificial intelligence to assessment. Charlotte, NC: Information Age Publisher. Cheng, J. (2018, September 2-6). Real-time scoring of an oral reading assessment on mobile devices[Paper presentation]. Interspeech, Hyderabad, India.
