An Approach to Backward Chaining in Xcerpt

Xcerpt is a declarative pattern based query and transformation language for semistructured data that like xml that bases on concepts and techniques of logic programming. Xcerpt uses both forward and backward chaining for program evaluation. While forward chaining is used for operations like view materialization, backward chaining is used in querying databases such as the web. This thesis presents several calculi that illustrate a possible formalization of the chaining process with respect to the the consequences of the nonstandard unification concept of Xcerpt as well as the special Xcerpt constructs all and some. These constructs are best comparable to the bagof predicate of Prolog. While leaning on positive analytic resolution calculi in the style of SLD, this thesis emphasizes the shortcomings of these classical calculi from logic programming for the evaluation of programs for semistructured data and hence the web, and why a constraint solving and nonlinear approach fits better with Xcerpt’s way of backward and forward chaining on semistructured data.