Many　people　use　the　web　as　the　main　information　source　in　their　daily　lives.　However,　most　web　pages　contain　non-information　components,　such　as　site　bars,　footers　and　ads,　etc.,　which　make　it　complicated　to　extract　text　from　the　original　HTML　documents.　Because　of　the　high　human　intervention　and　the　low　results　extraction　quality,　although　the　web　text　extraction　techniques　have　been　developed,　the　popularization　and　efficiency　of　the　usage　still　need　to　be　solved.　In　this　paper,　we　proposed　a　maximum　subsequence　segmentation　(MSS)　algorithm　and　discussed　its　application　in　the　domain　of　news　web　sites.　Differing　from　the　tree　structure　analysis　and　VIPS,　the　algorithm　divided　the　web　into　text　segmentation　and　label　segmentation.　Experiment　shows　that　the　MSS　algorithm　achieves　93.73%　accuracy　over　2000　news　pages　from　5　different　news　sites　and　the　efficiency　is　much　faster　than　DOM-based　using　same　dataset.　©　2013　IEEE.