Fenet Guyassa

Class of 2024

A non-experimental survey research design (also known as a questionnaire study) was conducted, but was originally intended to incorporate both components of quantitative and qualitative data--otherwise known as mixed methods--but was then narrowed down to a single quantitative approach. 

This research would be categorized as a create project since the pronunciation program was built. There is no precedent model with this exact replica. 

 Before carrying out this investigation, it was hypothesized that it is indeed possible to create an application that uses speech-to-text as a tool to help Ethiopians of all American English backgrounds. Even though this research is only applied to the Ethiopian teen and adult population, the researcher anticipates that the evidence generated by this study might be able to draw a connection between practice spent on the pronunciation program and improved pronunciation of a given word in English for younger individuals as well. 

Phonetics: Phonetics is a foundational branch, impacting other branches, in the study of linguistics [2]. The branch of phonetics refers to how speech sounds are produced as well as the acoustic properties and qualities of human speech [3]. Although there are three major concentrations within the field of phonetics, this paper focuses more heavily on acoustic phonetics. 

Speech Recognition: Speech recognition is the conversion of sound waves being translated into electrical signals that are processed to binary code by a machine [2]. Speech recognition can be considered an interdisciplinary niche of linguistics and modern day technology. 

Several major limitations may come to mind when considering what could have impacted the final results. This includes OS restrictions, inaccurate conversion from IPA to English, and different survey biases. 

In the programming, manual conversions from IPA to English were made so standard users could easily comprehend the speech sounds written in the form of text on screen. 

As mentioned in the Materials and Methods section, the program was available on nearly any Windows OS...without cross-platform access, confounding variables related to the characteristics that users of a particular OS share could have influenced the results. 

Finally, the survey was posted for about 1 to 2 weeks and collected only 16 responses. This does not meet the threshold of 30 to satisfy the Central Limit Theorem so no causal relationships can be drawn from these results. Convenience sampling is additional bias since respondents were contacted through personal connections. 

 Studying this demographic of the human population through two connected lenses, speech technology and phonetics, has not been tackled before and brings forth new implications. There are several ways that these findings can be applied in the real world, but a particularly important instance is education. 

There are roughly 305,800 Ethiopians in the United States, which was previously determined to be one of the largest immigrant demographics from Africa [4]. The pronunciation program can serve as a useful tool for thousands of Ethiopians who are classified as English as a foreign language (EFL) students in a classroom setting. 

It can be deduced that phoneme-based feedback from a speech-to-text system has shown to improve English word pronunciations of Ethiopian individuals. There are various possible applications of this research as well as bias and other limitations to note, however, it leads the way for future inquiries in technology and human languages.