Automatic Web Document Restructuring Based on Visual Information Analysis

document restructuring, page analysis, page segmentation, block importance

Many documents available on the current web have quite a complex structure that allows to present various kinds of information. Apart from the main content, the documents usually contain headers and footers, navigation sections and other types of additional information. For many applications such as document indexing or browsing on special devices, it is desirable that the main document information should precede the additional information in the underlying HTML code. In this paper, we propose a method of document preprocessing that automatically restructures the document code according to this criteria. Our method is based on rendered document analysis. A page segmentation algorithm is used for detecting the basic blocks on the page and the relevance of the individual parts is estimated from the visual properties of the text content.