Using Category Information for Relationship Exploration in Textual Data

Abstract In the comprehension of textual data, it is critical for people to perceive relationships between topics. This work explores two approaches that use text categorizations to reveal underlying relationships: the Overlap approach, which visualizes overlaps between categories, and the Search approach, which shows topical search results in the context of categories. The effectiveness of these approaches is tested using various types of relationship questions. Our results show that the Overlap approach improves users' performances in relationship exploration tasks. Conversely, the Search approach did not show the same effectiveness, primarily due to the Vocabulary Problem. Design implications are drawn from the experiment.