A New Method Supporting Qualitative Data Analysis Through Prompt Generation for Inductive Coding
Fengxiang Zhao, Fan Yu, Yi Shang · 2024
Recent advances in Large Language Models (LLMs) have revolutionized numerous fields, including Qualitative Data Analysis (QDA). This paper introduces a novel method, ArGUMENT2CODE (A2C), designed to leverage the capabilities of LLMs for enhancing the QDA process, particularly focusing on the inductive coding aspect. A2C sets itself apart from conventional automated coding tools by initiating a two-stage fine-tuned LLM process adept at navigating the complex landscape of qualitative data. This innovative method starts with the identification of coding cues hidden within the textual data, which are then refined into targeted prompts. These prompts are then used for guiding the inductive coding process, facilitating the generation of a rich, actionable codebook that offers expansive coverage of analytical perspectives. Our experimentation reveals that A2C not only successfully generates a pertinent and insightful codebook consistently but also significantly outperforms all other existing methods.