HierarchicalTopics: Visually Exploring Large Text Collections Using Topic Hierarchies

摘要: Analyzing large textual collections has become increasingly challenging given the size of data available and rate that more is being generated. Topic-based text summarization methods coupled with interactive visualizations have presented promising approaches to address challenge analyzing corpora. As corpora vocabulary grow larger, topics need be generated in order capture meaningful latent themes nuances However, it difficult for most current topic-based represent number without cluttered or illegible. To facilitate representation navigation a topics, we propose visual analytics system - HierarchicalTopic (HT). HT integrates computational algorithm, Topic Rose Tree, an interface. The Tree constructs topic hierarchy based on list topics. interface designed present content as well temporal evolution hierarchical fashion. User interactions are provided users make changes their mental model space. qualitatively evaluate HT, case study showcases how HierarchicalTopics aid expert making sense discovering interesting patterns groups. We also conducted user quantitatively effect structure. results reveal leads faster identification relevant solicited feedback during experiments incorporated some suggestions into version HierarchicalTopics.