From Content-based to Connectivity-based Search on the World Wide Web—A Journey Trail

Abstract

Searching the World Wide Web is an activity whose importance cannot be emphasized any further. While, searching the web, a ranked list or <i style="mso-bidi-font-style:normal">all<span style="mso-bidi-font-style: italic">-and-only </span></i>relevant document is desired, in response to a query. The web search efforts started with the application or classical content-based information retrieval technique' to the web. However, owing to commercial pressures,<i> spamming </i>became a major threat to the success or content -based technique. The work then diverted to exploit the hypertextual structure or the web. The number as well a quality or hyperlinks to a page is taken to be a measure of <i style="mso-bidi-font-style:normal">author<span style="mso-bidi-font-style:italic">ity </span></i>of that page. A page is more authoritative if many pages point to it. or alternatively, if some high-authority pages point to it. Such an authoritative value may directly be taken as a basis for ranking of several given pages. The approach help in fighting the <i>spamming </i>problem as the importance or a given page is now gauged not by the content of that very page <i>alone, b</i>ut rather by the (hypertextual) contents or other pages on the web. It gives an unbiased factor to the ranking procedure. Hypertextual structure has proved useful in finding the related pages. The quest for related page is nothing but search-by-example. Instead of issuing a query, we come up with a sample page and search for pages, similar to it. It helps in situations where formulation or a suitable query is difficult, particularly when the information space is unfamiliar to the user. If one models the web as a directed graph, with the web pages being the nodes and the hyperlinks between them bring the edges or the model graph, then starting from a sample page, the breadth –first search to a few levels would give a set or pages that are related to the sample page.