Poster
Authors: Hannah Kim, Jaegul Choo, Alex Endert, Haesun Park
Abstract: We present a visual analytics system for large-scale document retrieval tasks with high recall where any missing relevant documents can be critical. Our system utilizes a novel user-driven topic modeling called targeted topic modeling, a variant of nonnegative matrix factorization (NMF). Our system visualizes a topic summary in a treemap form and lets users keep relevant topics and incrementally remove uninteresting topics in our treemap view without losing potentially relevant documents.
