Evidence and Assessing Verbatim Transcription Quality of Black English and Standard American English
Abstract
This chapter explores the accuracy of private transcription services when transcribing Black English and Standard American English. Courts and lawyers in the US regularly rely on transcripts from such services, but third-party verification of their accuracy, especially with respect to their quality when faced with nonstandard language varieties, is lacking. This study draws on experimental methods to contrast the quality of transcription services offered by transcribers and AI route. The quantitative results show that transcription by humans and AI resulted in more mistakes when transcribing Black English than when transcribing Standard American English. Furthermore, a qualitative analysis reveals that these mistakes often changed or obscured meaning in legally relevant ways. If these results are generalizable, many transcripts currently in circulation, and crucial both to justice at the trial level and appellate review, contain disproportionately more legally important mistakes for Black English speakers. Given that Black English speakers are a highly overrepresented population in the US criminal system, the chapter proposes ways of redressing transcription shortcomings.
Fields of Study & Subjects
Legal processes and jurisprudence
Linguistic Variation and Morphology
Artificial Intelligence in Law
Publication details
- Journal / source
- Cambridge University Press eBooks
- Pages
- 80–96
- Publication date
- 19 February 2026
- Language
- EN
- Document type
- Book chapter