Page query language generation for structural extraction

The information on the Web is usually fabricated to be understandable by human users rather than machines. It´s not easy to automatically catalogue and extract the Web information solely with a software agent. Based on these observations, we present an approach that uses human guided operations to automatically generate a PQL query, a SQL like query language focusing on Web pages, to extract the interested information fragments on Web pages. The PQL query uses XPath expressions to locating the target HTML nodes. We develop a K-Medoid clustering algorithm to process PQL queries to generate the structural extractions. The extracted information is structured as a relational table (in CSV format) which can be manipulated smoothly with spreadsheet software or a relational DBMS system.