A New Method Using LLMs for Keypoints Generation in Qualitative Data Analysis

Abstract

Qualitative data analysis (QDA) is useful for identifying patterns, themes, and relationships among data. In this paper, we propose a new method that uses large language models (LLMs), such as GPT-based Models, to improve QDA, in Ecological Momentary Assessment (EMA) studies as an example, by automating keypoints extraction and relevance evaluation. Experimental results on the IBM-ArgKP-2021 dataset show improved performance of the new method over existing work, achieving higher accuracy while reducing time and effort in the coding process of QDA, and demonstrate the effectiveness of our proposed method in various application settings.