Object Extraction Based on Spatial-Relation of Entities from the World Wide Web

The spatial distance between components within one object is always less than that between different objects in Web pages. A novel method of object extraction from the World Wide Web is reported. This proposed method considers mainly the layout characteristic of Web contents and is independent of underlying documentation representation such as HTML code. The method is based on document object model (DOM) to obtain the bounding-box of various kinds of Web information such as image, text or link. Then the distance of adjacent components is computed to get the spatial relation. Finally, all the Web information components of the same object can be integrated. Experiments showed that the proposed method could work well even when the HTML structure was far different from layout structure, and the experimental results are quite encouraging.