A study on using two-phase conditional random fields for query interface segmentation

摘要: Recently, the Web has been rapidly "deepened" by many searchable databases online, where data are hidden behind query interfaces. Automatic processing of a interface is must to access invisible contents deep Web. This entails automatic segmentation, i.e., task grouping related components an together. The segmentation divided into two steps: component labeling and grouping. In this paper we present new approach perform using two-phase Conditional Random Fields (CRFs). At first phase, one CRFs model used tag each with semantic label (attribute-name, operator, operand or other); at second another create groups components. Experiments show that our yields high accuracy.