An Evaluation of Expedited Transcription Methods for School-Age Children’s Narrative Language: Automated Speech Recognition & Real-Time Transcription

An Evaluation of Expedited Transcription Methods for School-Age Children’s Narrative Language: Automated Speech Recognition & Real-Time Transcription

Logan, UT

In this study the clinical utility of several expedited transcription methods were evaluated for narrative language samples elicited from school-age children (7;5-11;10) with developmental language disorder (DLD). Transcription methods included real-time transcription (RTT) from both school-based speech language pathologists (SLPs) and trained transcribers (TTs), and automatic speech recognition (ASR). Each method transcribed the same 42 language samples, which were compared to a reference corpus produced using traditional transcription. The accuracy of each method was evaluated using a cross-sectional multilevel model. Results indicated that ASR was significantly more accurate than RTT from both SLPs and TTs. The reliability of scores produced using each transcription method against the reference corpus was evaluated on five quantitative metrics from a SALT Standard Measures Report. Pearson correlations revealed ASR had the highest reliability across all 5 quantitative metrics. These findings collectively indicated that ASR had greater clinical utility than RTT for use with narrative language samples of school-age children with DLD. This research was funded by the Graduate Research; Creative Opportunities Grant awarded by Utah State University. Presentation Time: Wednesday, 10-11 a.m.