In　order　to　provide　basic　data　for　improving　the　intellectual　property　early　warming　capacity　and　the　competitiveness　of　high-tech　industries　of　Beijing,　by　searching　the　database　of　the　United　States　Patent　and　Trademark　Office,　patent　information　in　the　form　of　dynamic　pages　can　be　gotten.　Based　on　XML　related　technology,　a　method　to　extract　and　store　patent　information　in　local　relational　database　is　put　forward　in　this　paper.　The　web　pages　are　filtered　by　regular　expression　matching,　and　then　the　document　object　models　of　the　pages　are　cleaned.　Finally　the　patent　information　is　extracted　by　XSLT　matching　and　stored　to　relation　database　by　object　mapping.　The　prototype　of　the　patent　extraction　system　is　designed　and　implemented,　which　has　a　high　recall　rate　and　precision　rate.