Where to Start Browsing the Web?

Both human users and crawlers face the problem of finding good start pages to explore some topic. We show how to assist in qualifying pages as start nodes by link-based ranking algorithms. We introduce a class of hub ranking methods based on counting the short search paths of the Web. Somewhat surprisingly, the Page Rank scores computed on the reversed Web graph turn out to be a special case of our class of rank functions. Besides query based examples, we propose graph based techniques to evaluate the performance of the introduced ranking algorithms. Centrality analysis experiments show that a small portion of Web pages induced by the top ranked pages dominates the Web in the sense that other pages can be accessed from them within a few clicks on the average; furthermore the removal of such nodes destroys the connectivity of the Web graph rapidly. By calculating the dominations and connectivity decay we compare and analyze the proposed ranking algorithms without the need of human interaction solely from the structure of the Web. Apart from ranking algorithms, the existence of central pages is interesting in its own right, providing a deeper insight to the Small World property of the Web graph.