ChronCN: Visualizing Topic Model Output with Chronological Chain Networks

This paper presents Chronological Chain Network (ChronCN), a network-based visualization framework designed to enhance the presentation and interpretation of topic model outputs. ChronCN addresses two challenges: (1) effectively visualizing essential informational dimensions of topic models, including temporal variations in topic prevalence, and (2) capturing the sequential dynamics of user engagement with different topics. To assess ChronCN, we employ a twofold evaluation strategy. First, we conduct a comparative analysis of existing visualization methods for topic modeling and position ChronCN within the current landscape. Second, we perform a preliminary user study to examine the users’ ability to identify and comprehend the informational elements depicted in various visualizations. Our findings indicate that users recognize the richness of informational content conveyed by ChronCN and are generally successful in accurately interpreting the visualization. Furthermore, ChronCN’s network-based design facilitates the visualization and simulation of potential user engagement paths across topics.