Each node in the XML tree is labeled with the identifier which consists of path information and sibling information. We constructed two types of indexes, one is for handling structural searches (STB-tree) and the other is for handling full-text searches (COB-tree). The search keys of the COB-tree are pairs of text fragments and the identifiers of the leaf nodes that contain the text, whereas the search keys of the STB-tree are the node identifiers.
STB-tree can handle queries which contain "//" such as "//title" efficiently because the search keys of STB-tree are ordered by the reverse path information. COB-tree keeps text fragments based on suffix texts. We also considered a compression technique for COB-tree.
Overview of XICS.