World Englishes and Whisper: Testing AI and Pedagogically Reframing Oral History Transcription Workflow

At New York University Abu Dhabi, a global liberal arts and science research campus, the Library’s University Archives is the home and facilitator of the community’s participatory oral history project, the NYUAD Memory Project; a project developed and inspired by Memory Projects around the world. Initiated in the Fall of 2019, and largely a remote interviewing project since spring 2020, our modest project has resulted in the recording of 95 interviews with a variety of community members, students, faculty, staff, and alumni. These interviews have all been conducted in English, the official teaching language of the campus; but the knowledge shared through memory and story represents a broad diversity of narrators' English speaking backgrounds: NYUAD hosts a student body representing over 120 countries and over 100 languages, and its faculty and staff represent over 50 nationalities. Within this context, University Archivist Lauren Kata interviewer and archivist for the NYUAD Memory Project – has begun partnering with NYUAD academic writing instructors Aieshah Arif and Sweta Kumari to consider and assess AI transcription tools and human transcription editing workflows within the framework of “World Englishes,” as first discussed by Braj B. Kachru (1982). As writing professionals and researchers, Arif and Kumari have applied Kachru’s and others’ World Englishes paradigm to explore the limitations of solely using “standard' varieties of English in their language instruction. This has led them to create space for diverse backgrounds and varieties of English, in order to foster inclusivity in the teaching and learning of the English language. This talk will discuss our collaboration still in its early stages – and how Lauren is taking the occasion of testing Whisper's performance to also reframe and evolve transcription “clean up” protocols within overall access workflows in the Archives. Applying the same inclusive writing pedagogies used in the University's Writing Center is leading to a reconsideration of editing methods and standards for speech-to-text English transcripts, i.e., the "next steps" that follow the use of AI tools, which involve the training and contributions of undergraduate student assistants. Examples and patterns encountered will be included in the presentation, such as Whisper's performance in generating transcripts of multiple English accents and varieties at NYUAD, and English expressions and phrases we’ve identified to be retained "as is" for authenticity and accuracy. In participating in this symposium, we hope for meaningful feedback from the larger community and others testing and encountering similar issues.