Abstract: Generating wrapper efficiently for extracting Web data has broad application prospect, but is also a difficult problem that is not yet solved efficiently till now. To tackle this problem, a pattern algebra for HTML documents is introduced, which includes key concepts, such as consistent pattern set, and the addition operation of pattern, and based on it a new approach to extract Web information is presented. It induces the consistent pattern set which represents identifying rules of each attribute by exploring the whole samples, and then extracts data by the consistent pattern set with multiple patterns. It can apply Web pages with tabular structure, in which there are missing attributes or attributes with multiple values or different order and hierarchical structure, and has been validated experimentally in the prototype.