Deep Web结果页面大多由网站根据请求从后台数据库读取数据并动态填充到通用模板而生成的。研究如何从一系列同模板生成的页面中生成该模板,并利用模板自动抽取数据。给出了模板生成问题的形式化描述、提出了一种新颖的模板生成方法,利用生成的模板从实例网页中抽取数据。与现有方法相比,该方法适用于列表页面和详细页面两种类型网页。通过在多个领域站点上实验,说明新方法在不降低准确率的情况下能大大提高召回率。
Most Deep Web result pages are dynamically generated using a common template populated with data from databases by user' s request. The research was to automatically generate template behind these template-generated Web pages and used the generated template to extract embedded data automatically. Folvaalized the template generation problem. This paper presented a novel template generation method and used the generated templates to extract data from instance pages. Comparing with existing research, this method was applicable for both list pages and detail pages. By the compare on several domains, the experiment results indicate that the novel method greatly improves recall on the base of high accuracy.