Hyperbolic Tree of Similar Pages

6. Results
Our project allows the user to see the similarity relationship between different search results. This provides a surprising insight about the kinds of results search engines provide. This is best illustrated through an example. We refer figure 2 again: this tree was created using a query on “dog”. We see from the figure that there is a large number of dotted lines between the nodes of the topmost branch of the tree (stemming from dogpile.com). This shows us that all the results in that branch are strongly “similar” to one another. By looking at the different nodes of that part of the tree, we quickly realize that they consist of search engines.
The other two top level search results also have some backward links, although less so. The tree thus tells the user some information about the search results that could not have been obtained through, or at least, would not have been as obvious in, a traditional search engine interface.
Suppose that the user finds dog.com to be the most interesting and wants to expand on that part of the tree. He types 3, the ID of dog.com, in the rightmost textbox and clicks “Make Root”, making dog.com the root of the tree, as shown in Fig 3. Now we observe that there are a lot more cross links throughout the tree, suggesting that the search results shown on the screen now are more similar to each other.


Figure 3

Now we consider another example. Searching for “Stanford University” yields the results as shown in Figure 4. Notice that there are a lot more dotted links in this example than in the previous one. This suggests that all the results shown are closely related to each other. This may allow the user to infer more information about the search results. One possible interpretation is that the search engine was able to find only one topic that maches the query. This may in turn suggests that the user can place greater confidence in the search results.

Figure 4