Felicidades to Professor Justin Davidson (Spanish and Portuguese) and undergraduate senior Andres Sanchez (Department of Linguistics), who were awarded a Latinx Research Center Faculty-Mentored Undergraduate Research Fellowship for Fall 2022, Spring 2023, and Summer 2023! The project, entitled "Accented Speech Recognition in Digital Voice Assistants: Addressing Spanish-Accented English," aims to expand accessibility, and ultimately inclusivity, when it comes to the utility of Automated Speech Recognition (ASR) tools by the Latinx population. ASR technology, which among other applications, forms the basis for digital voice assistants like Apple’s Siri, Amazon Alexa, and Google Assistant, has traditionally been developed on the basis of speech patterns of monolingual, Caucasian English speakers. Beyond calling for a paradigm shift in the technological sector that better foregrounds minority populations and Spanish-English bilingualism in the US, this research stands to directly afford more equal access to ASR tools by the Latinx community. More specifically, in this project we will compare the accuracy (via Word Error Rate) of the aforementioned digital voice assistants between Latinx Spanish-English bilinguals and Caucasian monolingual English speakers from the California Bay Area, and subsequently will adopt select machine learning solutions to address performance disparities between the groups. Combining methodologies from the Humanities, Social Sciences, and Computer Science, we believe this project will inspire continued interdisciplinary collaboration and highlight the crucial need for ASR technology to be developed with minority populations, and in particular the Latinx community, in mind.
Felicidades to Professor Justin Davidson (Spanish and Portuguese) and undergraduate senior Andres Sanchez (Department of Linguistics), who were awarded a Latinx Research Center Faculty-Mentored Undergraduate Research Fellowship for Fall 2022, Spring 2023, and Summer 2023! The project, entitled "Accented Speech Recognition in Digital Voice Assistants: Addressing Spanish-Accented English," aims to expand accessibility, and ultimately inclusivity, when it comes to the utility of Automated Speech Recognition (ASR) tools by the Latinx population. ASR technology, which among other applications, forms the basis for digital voice assistants like Apple’s Siri, Amazon Alexa, and Google Assistant, has traditionally been developed on the basis of speech patterns of monolingual, Caucasian English speakers. Beyond calling for a paradigm shift in the technological sector that better foregrounds minority populations and Spanish-English bilingualism in the US, this research stands to directly afford more equal access to ASR tools by the Latinx community. More specifically, in this project we will compare the accuracy (via Word Error Rate) of the aforementioned digital voice assistants between Latinx Spanish-English bilinguals and Caucasian monolingual English speakers from the California Bay Area, and subsequently will adopt select machine learning solutions to address performance disparities between the groups. Combining methodologies from the Humanities, Social Sciences, and Computer Science, we believe this project will inspire continued interdisciplinary collaboration and highlight the crucial need for ASR technology to be developed with minority populations, and in particular the Latinx community, in mind.