8. Conclusion We have proposed and implemented a new graphical representation of the similarity relationships between web search results. In doing so, we have found that there is additional information in search results which are not being used because of the lack of an appropriate interface. By providing the user with the tools to visualize the relationship between search results, we feel that we can enable the user to find a group of relevant search results more readily.
While both similarity relationships and hyperbolic trees have been studied in the past, to the best of our knowledge, our use of hyperbolic trees to represent similarity relationships is new. We have also proposed a novel method of reducing the clutter imposed by cycles in drawing hyperbolic trees by using lines of differing weights based the length of the cycles.
9. Future Work In our project, we made assumptions about the similarity relationships. The validity of these assumptions need to be studied. In particular, our tree representation of the similarity relationship is not strictly valid if the reflexivity assumption fails. Under such circumstances, other methods of modifying the tree may be more appropriate.
Our interface can also be improved. Due to time constraints, we were only able to build a simple interface to illustrate the concepts we have proposed. A more usable interface would allow the user to click directly on the tree to manipulate it. The user should also be able to open up the web results shown in the tree by directly clicking on those nodes.
References 1. Inxight Software. http://www.inxight.com/news/usability_study.html 2. Visualizing the Structure of the World Wide Web in 3D hyperbolic space. Tamara Munzner and Paul Burchard. ACM SIGGRAPH, New York, 1995.
3. Juan C. Dürsteler. Info@Vis. http://www.kartoo.fr/en/press/2002/press_aout/infovis.htm