Gerald Benoit, Dept of Computer Science & Grad School of Library & Information Science
Simmons College
Gerald Benoit reported for us the rationale for a 3D-based interface for navigating the results from a search query of an Art History database of images. He reviewed other 3D interfaces, reported on the user-study of his own system (which, he admitted, was a rejection of their 3D approach), and discussed the test subjects interest in “ownership” of the catalogue and its meta-data.
A draft of his paper, “Interactive 3D Visual Retrieval for Art History Education”, is available at his website.
Benoit’s proposed at least two reasons for exploring 3D interfaces. First, is that a result-set can include relevance and clustering (e.g. associating one painting with another) data that is difficult to expose the user to. By representing the result-set in 3D space, the designers are afforded x,y, and z-axises as distinct dimensions for representing this meta-data. Secondly, Benoit suggested that a three dimensional space might be easier for certain types of people to navigate. He cited HCI research on “affect” as well as empirical work that differentiated how men and women navigate through space or how people from different disciplines (e.g. biologists) are more willing to explore novel interfaces.
Benoit reviewed a few 3D visualization systems (e.g. ArtSTOR and Greenstone). A particular criticism that he leveraged is that many of the visualization techniques: color-choices, object representation, different axises were meaningless to the user. This theme of producing a catalogue that the user has a sense of ownership over would carry over into his analysis of his own 3D retrieval system.
Next, Benoit programmed and user-tested a 3D image retrieval system for Art Historians. Among the different findings, it is safe to say that the users generally rejected the 3D approach. The responses that drew Benoit’s attention were the curiosity and requests that they had about: how to group their own findings (e.g. like a shopping cart) and why the IR system included certain images in the result-sets for their queries. Furthermore, he found a generational difference in that the younger generation showed very little interest in the hierarchical representation of the object. Instead, they were more interested in having access to high-quality images and data to use the catalogue meta-data as a resource (e.g. as a teaching aid).
In response, Benoit has been working with the Boston Public Library (and other institutions) in producing an image catalogue, named the AuroraDL, with enriched records that supports functionality that allows users to “own” their catalogue. The system is redesigned to permit users to create their own groupings, to annotate existing entries, and relabel others. User contributions are vetting through three stages: Expert, Curatorial, and Student. He expects to produce an integrated solution such that searchers do not have to export their findings and re-import them into another setting to continue working (a software limitation that distinguishes between “searching” and other tasks, such as composing an art lecture). He anticipates that the technical strengths of this project, combined with full-text data by user-contributors, will expand the computational possibilities for data-mining, automated cataloguing, and foreign-language translations.
