QuAD: Deep-Learning Assisted Qualitative Data Analysis with Affinity Diagrams

BSTRACT Affinity diagramming is an effective and efficient method for forming nuanced interpretations of wide-ranging, unstructured qualitative data; however, this method does not scale well to large data sets. We propose a novel affinity diagramming system, called Qualitative Affinity Diagrammer (QuAD) that leverages computer-generated suggestions using deep learning to address the scalability of the diagramming process. QuAD features automatic grouping suggestions to jump-start the affinity diagramming process and provides grouping suggestions throughout the diagramming process to reduce sifting through notes. In this paper, we present a prototype of QuAD that uses Bidirectional Encoder Representations from Transformers (BERT) and Girvan-Newman to generate grouping suggestions. This work is the first step towards creating a powerful tool for assisting in the analysis of large qualitative data sets in a variety of contexts, including human-computer interaction.