Efficiently enumerating results of keyword search

Various approaches for keyword search in different settings (e.g., relational databases, XML and the Web) actually deal with the problem of enumerating K-fragments. For a given set of keywords K, a K-fragment is a subtree T of the given data graph, such that T contains all the keywords of K and no proper subtree of T has this property. There are three types of K-fragments: rooted, undirected and strong. This paper describes the first provably efficient algorithms for enumerating K-fragments. Specifically, for all three types of K-fragments, algorithms are given for enumerating all K-fragments with polynomial delay. For rooted K-fragments and acyclic data graphs, an algorithm is given for enumerating with polynomial delay in the order of increasing weight (i.e., the ranked order), assuming that K is of a fixed size. Finally, an efficient algorithm is described for enumerating K-fragments in a heuristically ranked order.