Dom-based Automatic Webpage Scrolling In Mobile Search System

Nowadays the usage of Mobile phones are widely spread in our lifestyle, we usemobile phones as a camera, radio, walkman and even a tool to browse internet. Sincemostly web pages are created for Desktop computers, it is hard to navigate them withsmall screen devices. Hence there is a great interest in computer science to adopt suchpages with rich content into small screens of our mobile devices. But obviously, everyweb page got many different parts that do not have the equal importance to the end user.This fact makes us to consider some ways to identifying the most useful portion of thepage to the end user in the most convenient way that keeps the information integrated andno loss of information as well. Due to the fact that a mobile phone has a small screen andthe text-entry device is difficult to use, so they prefer to use mobile phones just once forperforming a search. The problem lays here. For a convenient use of phones for surfingthe net is related to their screen. There are two issues related to the screens of phoneswhich make it a drawback:1. the resolution is low, so some approaches proposed forzooming is not so useful.2. Long web content can’t be easily displayed in both verticaland horizontal way.The proposed architecture is written as a client-server fashion to meet therequirements of a mobile browser. In the client side, we open an instance of a Lobobrowser to prepare the user with a screen to search through the net. When a search requesthappens and the users choosing one of the results, the computation in server will bestarted. The server will first receive the URL and the keyword from client, and then sendthe URL to a function which parsing the HTML code and convert it into a DOMrepresentation. Onward traversing through the tree will be applied to get the nodesinformation and by stop word filtering and stemming, we will be sure we have the rootwords. Now we can store the root nodes in a temp file and calculate the TF-IDF value oneby one. The server will calculate the posterior probability for each part of the page. Atthis moment, we can take the keyword and will compare it with root nodes to find theright value for it. Then we calculate the posterior probability for the keyword andcompare it with each part’s value to find the best part as a candidate to layout to the useras the result. The result shows that the method behaves with good accuracy but it seems as long as the length of keyword grows, the accuracy gets lower, though the timespending for finding the most important part will not vary much.