English-to-IPA Transcription

Author

Riley Roberts, Utah State University

Date of Award

12-2021

Degree Type

Thesis

Degree Name

Departmental Honors

Department

Electrical and Computer Engineering

Abstract

The purpose of this project was to create a tool that could automate English-to-IPA (International Phonetic Alphabet) transcription. Research was done to determine what would be required to perform such a transcription. After researching and experimenting with existing tools, it was determined that developing the signal processing and Artificial Intelligence model portion of the application would be too intensive to successfully complete within the timeframe of this project.

The choice was made to develop an iOS application, with the Python library Allosaurus used to do the speech processing and as the Artificial Intelligence model. This model was then deployed in the cloud using AWS, and the two components, iOS and cloud, were then integrated to work together. The deployed cloud model was placed behind a secure API layer, requiring authentication for access, to ensure security and consistency of requests. The final application functions as intended and is capable of performing the desired transcriptions. The application follows conventional iOS user interface styling so-as to be conducive to easy pickup by the intended users, Speech Therapists.

The application is not yet ready for public release, due to the inability to meet some accuracy and security requirements that were originally specified. However, the researcher has plans to continue the work beyond the scope of this project and presents ideas for how to solve the current problems the application is facing. Additionally, the researcher discusses ethical considerations of the project, and supplies a User Guide with non-technical instructions for operating the application.

Recommended Citation

Roberts, Riley, "English-to-IPA Transcription" (2021). Undergraduate Honors Capstone Projects. 920.
https://digitalcommons.usu.edu/honors/920