Nowadays more and more Web sites generate Web pages containing client-side scripts such as JavaScript and Flash instead of ordinary static HTML pages. These scripts create dynamic HTML pages and provide modern interfaces to Web applications such as AJAX applications. Unfortunately, for partial information extraction of Web pages, existing methods cannot extract dynamic Web contents.

In this paper, we present a new partial information extraction method which can deal with not only static Web contents but also dynamic Web contents created by client-side scripts. As applications, we present personal mashup construction examples based on our extraction method. Our implementation shows that our extraction method is efficiently applicable to various types of Web sites such as news sites, country profile sites and weather sites.