User Profile-based Reference Points in Information Visualisation

Chaomei Chen1and John Davies2

1 Department of Information Systems & Computing Brunel University
Uxbridge
UB8 3PH, UK

2 Knowledge Management Research BT Labs
Ipswich
IP5 3RE, UK

ABSTRACT
This paper briefly describes the design of a novel user interface for exploiting documents accumulated in an information filtering and sharing environment. In addition to visualising inter-document relationships, the visual user interface reveals the interconnectivity between user profiles and documents. The work currently focuses on the role of user profiles based on the notion of reference points in order to provide additional heuristics for information sharing.

KEYNOTE ADDRESSES

PANELS

Keywords:

Visual user interface, information visualisation, reference points.

IMAGE imgs/HCI98CC23.gif

INTRODUCTION
The exponential growth of widely accessible information in modern society highlights the need for efficient information filtering and sharing. Information filtering techniques are usually based on the notion of user profiles in order to estimate the relevance of information to a particular person.
Jasper is an information filtering and sharing system (Davies et al., 1995). It maintains a growing collection of annotated reference links to documents on the World Wide Web (WWW). Currently, the interconnectivity among these accumulated documents and user profiles is not readily available in Jasper. In this paper, we describe the design of a novel visual user interface in order to uncover the interconnectivity.
REFERENCE POINTS
Our design is based on the notion of reference points. This concept was originated in psychological studies of similarity data and spatial density (Krumhansl, 1978). The underlying principle is that geometric properties such as symmetry, perpendicularity, and parallelism are particularly useful in communicating graphical patterns. For example, people often focus on structural patterns such as stars, rings and spikes in a network representation. Reference points, conceptually or visually, play the role of a reference framework in which other points can be placed. In our work, we hypothesized that a number of star-shaped, profile-centred document clusters would emerge if the role of reference points was actively played by user profiles. Users would be able to share information more effectively based on the additional information provided by user profiles through the visual user interface. METHODS
We randomly sampled 127 documents and 11 user profiles from Jasper. The mixed collection was visualised within the Generalised Similarity Analysis (GSA) framework (Chen, 1997). First, we extracted and preserved only the most salient semantic relationships in order to reduce the complexity of the visualisation network. Second, we incorporated user profile-based reference points in order to improve the clarity of the visual user interface. Unique behavioural heuristics were applied to distinguish user profiles and documents so as to speed up the convergence of our self- organised clustering process. These emergent structures were derived without any prior knowledge of structural relationships. Additional structural cues are likely to result in more efficient results.

ORGANISATION OVERVIEWS

INDUSTRY DAY

DEMOS

VIDEOS

DOCTORAL CONSORTIUM

POSTERS

--20--

HCI'98 Conference Companion

RESULTS
The impact of user profile-based reference points can be seen in Figure 1. The left sub-figure shows the self-organised spatial layout without using the mechanism of reference points. The sub-figure in the middle shows the layout if the mechanism of reference points was utilised. In fact, the 11 user profiles, which merely make up 8% of the 138 nodes, were associated with 69% of the links in the network, whereas the remaining 127 documents, which make up 92% of the nodes, only shared 31% of the links. Reference points have clearly improved the clarity of the overall structure. Users now may track relevant documents based on their knowledge of their colleagues' expertise.

IMAGE imgs/HCI98CC26.gif

Figure 1:

The role of reference points: disabled (left), enabled (middle), and a close-up look at a cluster (right) (cube=profile; sphere=document).

CONCLUSION
In this paper, we have shown that the quality of information visualisation can be improved by incorporating user profile-based reference points. Preliminary results suggest that this technique is potentially useful for visual user interface design. More work is needed to the study of human factors in using this type of visual user interface so as to enable users to gather and share information more efficiently. ACKNOWLEDGMENTS
This work was supported in part by 1997 BT short term research fellowship.

REFERENCES
Chen, C. (1997), Structuring and visualising the WWW by Generalised Similarity
Analysis, in Proc. of Hypertext'97(Southampton, UK, April 1997), ACM Press,
pp. 177-186.
Davies, N. J., Weeks, R. & Revett, M. C. (1995), An information agent for WWW, in
Proc. of the 4th Int. Conf. on World-Wide Web (Boston, USA, Dec. 1995). Krumhansl, C. L. (1978), Concerning the applicability of geometric models to similar
data: The interrelationship between similarity and spatial density. Psychological
Review85, 5, 445-463.