Automated Analysis of Teacher-Child Language Interactions in Pre-K Classrooms

• In Event: AERA Roundtable Session Saturday 1:45 pm - Gold 4
  In Roundtable Session: Multimodal Literacies: Visual, Digital, and Oral Perspectives

Abstract

Oral language development in pre-kindergarten is a key predictor of later academic achievement, but researchers face challenges in capturing language interactions at scale. This study developed an automated speech processing system designed to analyze teacher-child interactions in pre-k classrooms. Participants included 18 children and 3 teachers. Our custom speech processing pipeline automatically transcribed recordings and identified conversational exchanges. Our system detected over 100 interactions, many of which included extended and balanced participation from teachers and children. Preliminary results suggest this method offers a scalable, data-rich alternative to traditional observation or hand-coding techniques. Findings have implications for identifying patterns in classroom language use and supporting the development of tools that promote equitable language growth in early education.

Authors

• Elizabeth Burke Hadley, University of South Florida - Tampa

• John Hansen, University of Texas - Dallas

• Jiamin Xie, University of Texas - Dallas

• Eunsook Kim, University of South Florida

• Dwight Irvin, University of Florida

• Sara A. Smith, University of Florida