An Analysis of Large Language Models for African American English Speaking Children’s Oral Language Assessment
Journal of Black Excellence in Engineering, Science, & Technology · 04.12.2023
DOI: 10.65465/pfzc5664 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Black Excellence in Engineering, Science, & Technology, online erschienen am 04.12.2023. Verfasst von Alexander Johnson und 5 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben.
Themen: Kinderkrankenpflege
Abstract (Originalsprache)
This paper evaluates the performance of widely-used open-source automatic speech recognition systems in transcribing primarily African American English-speaking children’s speech for educational applications. We investigate the performance of the Whisper, HuBERT, and Wav2Vec2 ASR systems as well as the capability of the transformer-based language model, BERT, for automatically grading the student’s oral responses to assessment prompts through use of the generated ASR transcripts. We achieve a 95% oral response scoring accuracy through the methods described. We also show a thorough analysis of ASR system performance over a diverse set of metrics going beyond the standard word error rate.
Zitieren
APA
Johnson, A., Chance, C., Stiemke, K., Veeramani, H., Alwan, A., & Shankar, N. B. (2023). An Analysis of Large Language Models for African American English Speaking Children’s Oral Language Assessment. Journal of Black Excellence in Engineering, Science, & Technology. https://doi.org/10.65465/pfzc5664
Vancouver
Johnson A, Chance C, Stiemke K, Veeramani H, Alwan A, Shankar NB. An Analysis of Large Language Models for African American English Speaking Children’s Oral Language Assessment. Journal of Black Excellence in Engineering, Science, & Technology. 2023. doi:10.65465/pfzc5664
BibTeX
@article{Johnson2023,
author = {Johnson, Alexander and Chance, Christina and Stiemke, Kaycee and Veeramani, Hariram and Alwan, Abeer and Shankar, Natarajan Balaji},
title = {An Analysis of Large Language Models for African American English Speaking Children’s Oral Language Assessment},
journal = {Journal of Black Excellence in Engineering, Science, & Technology},
year = {2023},
doi = {10.65465/pfzc5664}
}
EUVIMED