Intelligent Web Navigation

摘要

Virtual integration systems retrieve information according to the user's interest. This information is retrieved from several web applications, but it is presented to the user uniformly, in an online process. Therefore, response time is a significant factor. An essential part of any information retrieval system is navigation through pages. Usually web pages contain a high number of links, some of them leading to interesting information, but most of them having other purposes, like advertising or internal site navigation. Traditional crawlers follow every link in each page, in order to analyze the target page, and classify it as interesting or irrelevant. This means having to retrieve, analyze and classify thousands of pages for every single site, which is a costly task. This problem can be solved with the combination of a web page classifier, to distinguish between interesting and irrelevant pages, and a link classifier, which automatically identifies links leading to interesting pages. This kind of navigation is more efficient and has a lower cost than traditional crawlers. Moreover, navigation model is automatically extracted from the site, instead of being handcrafted, reducing the supervision from the user.