In this paper, we investigate the architecture for visual social navigation systems, especially those used in scientific literature. We identify the requirements for effective visualization of reputation and recommendation, two notions that underlie most s
Visualizing Reputation and Recommendation in Scientific
Literature
Pearl Pu
Human Computer Interaction Group Global Computing Center
Swiss Institute of Technology Lausanne
Punit Gupta
Indian Institute of Technology Guwahati
Guwahati-781039, Assam India. CH-1015 Ecublens, Switzerland
Pearl.pu@epfl.ch
ABSTRACT
In this paper, we investigate the architecture for visual social navigation systems, especially those used in scien-tific literature. We identify the requirements for effective visualization of reputation and recommendation, two no-tions that underlie most social navigation systems. We present several designs of visual emphasis techniques addressing the dimension and scales of reputation, and two sets of icons for effectively distinguishing recom-mended items. The results of two user studies are then discussed, which provided validation for our non-linear emphasis methods to represent reputation, and for choos-ing optimal icons for recommended items. Based on these findings, we present a re-design for the NEC Research Institute ResearchIndex (Citeseer), a concrete case of reputation based recommendation system.
KEYWORDS
Visual reputation and recommendation systems, informa-tion visualization, social navigation, emphasis techniques, scientific literature search.
INTRODUCTION
From maps etched on cave walls indicating shared knowl-edge of hunting trips in ancient times, to dog-eared books in libraries marking their popularity, or smoothed trails in forests indicating paths leading somewhere, social environments have enabled humans to accomplish
complex tasks based on information gathered collabora-tively. Such information is often presented in the forms of objects bearing signs of how they have been used by oth-ers, whether as maps, tear-and-wear signs on books, or footprints [10]. At the same time, such information tends to be very visual. Partly it is because of the navigational nature of user tasks in such social environments (hunters must
punitkg@yahoo.com
quickly go on hunting, library users must quickly decide which books to read), thus requiring a quick-to-perceive medium for communication and understanding. Partly it is due to the fact that traditional social spaces are physical and therefore intrinsically visual. Today, however, we go to the World Wide Web to find out where to dine, which books or CDs to buy, or where to go for vacations [1]. Digital spaces are replacing traditional social environ-ments, and employ mechanisms of collaborative filtering and recommendation systems to enable users to navigate and find useful information in these spaces [12]. Unfortu-nately, digital spaces do not age, nor are they physical in nature. Most information concerning reputation, and rec-ommendation are text based, forcing users to “read” rather than affording them to “see” information, thus making it hard for them to quickly perceive evidence, spot and retain trends, and patterns. “Read” and “see” differ in terms of the two ways in which the human visual system processes information, one being controlled and other automatic processing [4][8].
We are concerned with investigating an architecture for visual social navigation systems, where “social translu-cency” is mapped to visual representation and therefore augments users capabilities in their navigation skills. We aim at developing methods to effectively communi-cate reputation and recommendation information to achieve the following objectives:
1. provide a social space that resembles the multi-dimensional physical space in which we habit; 2. provide distinguishable and intuitive visual signs
for reputation, enabling searchers to quickly de-cide which items to pursue (decision making in navigation), without making decisions for them; 3. provide visual representation for recommended
items to enable users to “see” items in the right categories, rather than “read” them;
4. provide users with quick learning curves to un-derstand and retain the meanings of signs in such environments.
Mining and Visualizing Recommendation Spaces for ...literature, and describe how a priori domain ...scientific problem itself, Rice and coworkers, ...
Mining and Visualizing Recommendation Spaces for ...literature, and describe how a priori domain ...scientific problem itself, Rice and coworkers, ...
热门文档