An automatic Web wrapper for extracting information from Web sources, using clustering techniques

We present a fully automated system (wrapper) for extracting information from semistructured Web pages. The emerging need for such systems occurs due to the need for going beyond the concept of "human browsing" by automating the process of information retrieval enabling further utilization by targeted applications. The key idea in our novel system is to exploit the format of the information contained in the Web pages discovering the underlying structure and finally maps it to semantic relationships. In doing this we identify one section of the Web page as the one containing the useful information and we proceed in extracting semantic tokens contained in this section by using clustering techniques and other tools of statistical origin. Our innovation consists in building a system that can operate without human intervention or training and yet achieving excellent extraction precision and recall.