Web Page Element Classification Based on Visual Features

Abstract

When applying the traditional data mining methods to World Wide Web documents, the typical problem is that a normal web page contains a variety of information of different kinds in addition to its main content. This additional information such as navigation, advertisement or copyright notices negatively influences the results of the data mining methods as for example the content classification. In this paper, we present a method of interesting area detection in a web page. This method is inspired by an assumed human reader approach to this task. First, basic visual blocks are detected in the page and subsequently, the purpose of these blocks is guessed based on their visual appearance. We describe a page segmentation method used for the visual block detection, we propose a way of the block classification based on the visual features and finally, we provide an experimental evaluation of the method on real-world data.