Hidden Web databases maintain a collection of specialised documents, which are dynamically generated in response to users’ queries. However, the documents are generated by Web page templates, which contain information that is irrelevant to queries. This paper presents a Two-Phase Sampling (2PS) technique that detects templates and extracts query-related information from the sampled documents of a database. In the first phase, 2PS queries databases with terms contained in their search interface pages and the subsequently sampled documents. This process retrieves a required number of documents. In the second phase, 2PS detects Web page templates in the sampled documents in order to extract information relevant to queries. We test 2PS on a number of realworld Hidden Web databases. Experimental results demonstrate that 2PS effectively eliminates irrelevant information contained in Web page templates and generates terms and frequencies with improved accuracy.