P. Au, M. Carey, S. Sewraz, Y. Guo, and S. M. Rüger. New paradigms in information visualization. In Proceeding of SIGIR’2000, pages 307–309, Athens, Greece, 2000. ACM Press. [pdf]

———————

The starting assumption of the paper is that of information overload the users have to fight when dealing with results from a search engine. In the authors’ opinion the strategy should be to shift the user’s mental load  from these slower thought-intensive processes such as reading to faster perceptual processes such as pattern recognition in a visual display.

Instead of the classical ranking, the author suggest clustering the hit documents and make use of the obtained groups with interactive displays. The propose three new paradigms for information visualization: (1) the Sammon Cluster View; (2) the Tree-map focus + context approach; (3) the Radviz interactive visualization.

The first (1) method is a conversion of the high-dimensional cluster centroid vector to two dimensions which preserves the distance among the clusters. This is then mapped on the interface, providing a landscape for navigation. The distance between the circles in the interface is then the metaphor of the similarity of their respective clusters.

The Tree-map (2) clustering is a 2-d space filling approach that was borrowed from D. Shneidermann. It consist in a technique that devide the space of a quares in nested sub-squares which have dimensions and position which reflect the original clustering and similarity among the documents.

In the Radviz visualization, related words are initially arranged on a circle and connected with an invisible spring to each document the appear to be in. The documents are thus placed in a equilibium of positions between their related words and the center of the circle.

Using the first viz, the author showed a setting with an holistic view giving primarily information about a first-order cluster structure and inter-cluster relations. Using the second visualization, it was possible to show the second-order cluster structure. With the third approach they showed a solution where the user could participate in the clustering process, setting the priorities of the relevant words upon the visualization.

Tags: clustering, information retrieval, information visualization, pattern recognition