Setsuko Yokoyama is an Assistant Professor of Digital Humanities at Singapore University of Technology and Design. Her current book project is about a literary history of speech-to-text technology, which reimagines the contemporary discourse on fairness in automated speech recognition (ASR) systems. Drawing from archival records of American literary artists who negotiated questions of race, gender, class, disability status, and national origin as they rendered folk speech sounds to text, her work argues speech-to-text remediation as editorial labor and contests the current software engineering workflow that assumes ASR data labeling work as low-wage, menial labor. 

Accented DH: Localized Analytical Frameworks for Singlish Automated Speech Recognition 

In the field of Digital Humanities (DH), the interplay between local specificities and global legibility has always offered the opportunity to cross-examine positionalities of any given research project. While any efforts to articulate the positionality help researchers to be cognizant of their own cultural hegemonies, research conducted in Singapore still risks misrecognition of its social significance when it does not seemingly align with the Anglo-American intellectual discourse. This presentation introduces the “Accented DH” framework to address such a challenge of critical DH research and invites the audience members to consider its utility for humanistic inquiries of digital technologies. Using the analysis of a Singlish automated speech recognition (ASR) system as a case study, the presentation specifically deliberates on the epistemology of the “accented”—a mode of perception that is mediated by the dynamic constitution of being an accented speaker—and its potential to resist the dominating discourse on fairness in ASR while carving out a unique vantage point to develop more equitable ASR systems.