Abstract
Aim: This study explores the semantic similarities between qualitative research transcripts produced by artificial intelligence (AI) and those transcribed manually, with a particular focus on challenges encountered when working with multicultural participants in health science research who are non-native English speakers.
Method: The analysis is based on an audio file from one representative participant in a qualitative study involving 20 participants. It compares transcripts generated by a professional audio transcriptionist with those produced by two AI platforms, Otter.ai and Avidnote.
Results: Findings reveal that while AI transcription has advantages in speed and cost-effectiveness, it can struggle with speaker differentiation and punctuation accuracy, necessitating manual review. Both platforms faced challenges with cultural terminology and accented speech, but Avidnote showed better performance in word recognition and comprehension. Limitations were primarily in the transcription of te reo Māori.
Conclusion: The study highlights the critical role of culturally competent researchers in reviewing transcripts to ensure accuracy and clarity. These findings contribute to a deeper understanding of the benefits and limitations of AI transcription tools in qualitative health research, especially when working with linguistically and culturally diverse populations.