Improving Automated Scoring of Prosody Using Deep Learning Algorithm

• In Event: Research Blitz: Automated Scoring

Abstract

This study proposes and evaluates an improved approach for automated scoring for prosody in oral reading fluency assessment. This research explored different model structures by using deep learning with prosodic and frequency-related acoustic features to produce better performance especially in cross-domain validation, where phrases are assumed to be unknown.

Authors

• Kuo Wang, Southern Methodist University

• xin qiao, Southern Methodist University

• George Sammit, Southern Methodist University

• Eric C. Larson, Southern Methodist University

• Joseph F. T. Nese, UNIVERSITY OF OREGON

• Akhito Kamata, SOUTHERN METHODIST UNIVERSITY