SHIFTR: a user-directed, link-based system for ad hoc sensemaking of large heterogeneous data collections

ABSTRACT

We present a novel method and prototype system to help users make sense of and reorganize large amounts of heterogeneous information. Our work is grounded in theories of categorization from cognitive psychology and is designed for ad hoc sensemaking; that is, supporting people's shifting goals and flexible mental representations of concepts. Shiftr adapts a carefully chosen Belief Propagation algorithm from large-scale graph mining to efficiently assist users in interactively clustering information of arbitrary types. The system functions effectively with few human-labeled examples, and supports the use of both positive and negative examples. We demonstrate Shiftr's utility through sensemaking scenarios, one of which uses the DBLP bibliography dataset, which contains more than 1.7 million author-paper relationships.

INDEX TERMS

Primary Classification:
  H. Information Systems
  H.5 INFORMATION INTERFACES AND PRESENTATION (I.7)
      H.5.2 User Interfaces (D.2.2, H.1.2, I.3.6)
          Subjects: Interaction styles (e.g., commands, menus, forms, direct manipulation)

General Terms:
Design

Keywords:
heterogeneous, information retrieval, sensemaking