Date of Award

Spring 2021

Thesis Type

Open Access

Degree Name

Honors Bachelor of Arts

Department

Computer Science

Sponsor

Dan Myers

Committee Member

Valerie Summet

Committee Member

Jay Yellen

Abstract

Stuttering is a speech impediment that often requires speech therapy to curb the symptoms. In speech therapy, people who stutter (PWS) learn techniques that they can use to improve their fluency. PWS often practice their techniques extensively in order to maintain fluent speech. Many listen to audio recordings to practice where a single word or sentence is played on the recording and then there is a pause, giving the user a chance to say the word(s) to practice. This style of practice is not customizable and is repetitive since the contents do not change. Thus, we have developed an application for iPhone that uses a text-to-speech API to read single words and sentences to PWS, so that they can practice their techniques. Each practice mode is customizable in that the user can choose to practice certain sounds that they struggle with, and the app will respond by choosing words that start with the desired sound or generate sentences that contain several words that start with the desired sound. We generate sentences in two ways: an AI approach using recurrent neural networks and a “fill-in-the-blank” approach. When generating sentences, the goal is that the sentences are relatively short and contain simple words, so that the user does not struggle to repeat the word or sentence.

Share

COinS