Multi-stage modeling of HTML documents.

摘要

The goal of this thesis is to first give the reader an accurate picture of several models of both information discovery and extraction within the World Wide Web and how those two processes are becoming increasingly interrelated in overall information analysis. Furthermore, it will investigate how a sophisticated analysis of visual documents, such as those on the Web, is becoming increasingly important in both finding and understanding the context of document information. The thesis presents several problems within document analysis and then tries to approximate solutions to those problems in a general analysis framework, which is implemented in a prototype application. Finally, an instance of the framework is used to demonstrate its own practicality by accumulating statistics on features of web documents such as script and style usage that are only discovered by a deeper document analysis.