Interviews have long been a crucial data source in qualitative research. By analyzing interview transcripts, researchers can gain deep insights into participants' experiences, perspectives, and emotions. Among various analytical approaches, coding is one of the most commonly employed methods. However, current manual coding processes are not only time-consuming and labor-intensive but also prone to inconsistencies due to differing viewpoints among researchers. Furthermore, prolonged, high-intensity coding tasks can impose significant emotional and psychological burdens.To address these challenges, this study proposes a module based on Large Language Models (LLMs) that employs Content Analysis for meaningful segmentation of texts. By leveraging prompts and contextual conditions, the model is designed to interpret semantic nuances accurately and automatically match appropriate codes.To evaluate the usability of this approach, we applied it to interview transcripts from educational research and compared the automated coding results with those generated by professional researchers through manual coding. Cohen's Kappa coefficient was used to assess the consistency between the two. Preliminary findings indicate that LLM-based automated coding demonstrates strong agreement with manual coding conducted by experts, suggesting a shared perspective on coding. These results highlight the potential of LLMs in analyzing and coding qualitative interview data.

Authors:
Tzren-Ru Chou, National Taiwan Normal University, Taiwan
Chih Chang Yang, National Taiwan Normal University, Taiwan
Shu Wei Liu, National Taiwan University of Science and Technology, Taiwan

About the Presenter(s)
My name is Chih-Chang Yang, and I am currently serving as a Senior Technician at Taiwan National Open University, where I have worked for seventeen years. My current responsibilities include planning and recording digital learning course videos using

See this presentation on the full schedule – Saturday Schedule

Bookmark