Faculty mentor/PI email address
garwoods@rowan.edu
Is your research Teaching and Learning based?
1
Keywords
Lecture Capture, Lecture Recording, Automated Speech Recognition (ASR), Transcription, Accessibility, Undergraduate Medical Education
IRB or IACUC Protocol Number
PRO-2025-173
Date of Presentation
5-6-2026 12:00 AM
Poster Abstract
To increase the accessibility and utility of lecture capture recordings, schools may utilize automated speech recognition (ASR) to generate closed-captions and transcripts. But, how accurate are these systems? This study analyzed 50 ASR-generated transcripts using word error rate (WER) and further sought to determine factors that impact accuracy. Only 19 of 50 (38%) transcripts met the 99% accuracy standard required by the Americans with Disabilities Act (ADA). The lecturer's accent was the most significant factor affecting accuracy, though medical jargon also presented a challenge. These findings suggest system improvements and manual review are necessary to ensure students receive accurate information.
Disciplines
Educational Technology | Medical Education | Medicine and Health Sciences
Included in
Data-Driven Analysis of Lecture Capture Transcript Accuracy of the Biomedical Foundations Course using Automated Speech Recognition
To increase the accessibility and utility of lecture capture recordings, schools may utilize automated speech recognition (ASR) to generate closed-captions and transcripts. But, how accurate are these systems? This study analyzed 50 ASR-generated transcripts using word error rate (WER) and further sought to determine factors that impact accuracy. Only 19 of 50 (38%) transcripts met the 99% accuracy standard required by the Americans with Disabilities Act (ADA). The lecturer's accent was the most significant factor affecting accuracy, though medical jargon also presented a challenge. These findings suggest system improvements and manual review are necessary to ensure students receive accurate information.