Data visualization is a perennial challenge that offers unique opportunities in analysis and storytelling. Similar to how texts from social media posts and news headlines can be analyzed and visualized, it is possible to apply similar techniques with videoconferencing transcripts and journal manuscripts. This paper presents two case studies in applying visualization methods using text-based data. The first case study visualized videoconferencing transcripts between design professors and practitioners through human participants assigning and aggregating keywords to enrich the dataset. The second case study utilized automation to visualize abstracts published by select design journals from the past two years. These two approaches will be compared to discuss the advantages and challenges associated with algorithm-driven efficiency and bias in data visualization. The outcomes of these endeavors reveal how visualizing dialogues and manuscripts can unveil narratives and insights that might have been missed in other modalities. By extracting the top keywords in a body of abstracts and transcripts, it was possible to see the collective areas of expertise, trending research topics in design, as well as knowledge gaps in the venues where the source text was extracted from. Ultimately, these projects raise questions on the assumptions behind data visualization outcomes and processes. What are the limitations of visualizations for topic modeling? What are the advantages and disadvantages of relying on machine learning algorithms in topic modeling? And how can humans and algorithms work together to promote learning experiences? These are some of the major questions that are explored through this study.
Associate Professor, College of Design, University of Minnesota, Twin Cities, Minnesota, United States
Data Visualization, Topic Modeling, Algorithms, Graphic Design