Scalable precondition-aware service composition with SPARQL

Functional service composition builds a plan that fulfills some user-provided goal using available services. The literature describes several approaches for description of services and goals, with different levels of expressiveness, ranging from the ones based only on inputs and outputs to the more expressive logics-based solutions. The former provide better scalability and performance, while the latter allow for increased expressivity in preconditions and effects of both services and goals. The approach proposed in this paper aims to achieve a balance between expressivity and performance of functional service composition. To this end, a graph-based composition algorithm is extended to support preconditions and effects described with a small subset of SPARQL. Experiments compare the performance of this extended algorithm with two state-of-the-art algorithms that support preconditions and effects and demonstrate better scalability. In one case, a maximum speedup of 29 times was achieved for the problems that could be expressed with the SPARQL subset. In another, problems that a state-of-the-art algorithm could not solve after 5 minutes where solved in seconds.