Modeling Composition of Semantic Web Services

The objective of the proposed PhD research is to investigate and contribute in the area of rigorous/formal modeling of SWS (Semantic Web Service) composition, using a multi-layer approach that takes advantage of e.g. logic-based, process algebraic and petri-net based formalisms. Roughly, the composition problem for SWSs is (a) the problem of recognition of an appropriate integration pattern and (b) the problem of identification of suitable component SWSs that can fit at specified places in such a pattern. It is useful to observe that in the case of two services (or a service and a client) (a) can be rephrased as the problem of recognition of a cooperation pattern, while (b) is the problem of cooperability verification. Having made this observation, we can now point to the originality of approach in the proposed research. Indeed, discovery of a cooperation pattern can be semantically driven, focusing on the desired goal state of the composite service (which underlies the planning-based approaches to WS composition), or formal, focusing on the observable behavior patterns of the services (message exchange sequences). The latter is agnostic as to the WS intentions in the context of cooperation and it is referred to in the literature using the semi-formal notion of a WS “choreography” [10], [11]. As a rudimentary approach in the direction of formal choreography we mention [1], [2] (despite the fact that such a connection is not made in [1], [2]). This approach is further explored in [3], [4] (extending [5]). Essentially a planning-based approach to composition underlies OWL-S [9]. Similarly, a planning based approach essentially underlies the significant contribution in automating WS composition presented in [7]. To the best of our knowledge, there is no single piece of research work that attempts to combine these complementary, in our view, approaches to WS composition and it is precisely this that constitutes a core objective of this proposed research. It should be further pointed out that, addressing the composition problem is a task with mutual dependencies on the tasks of (1) service specification (e.g. the top ontology for semantic –ontologically engineered– web services) and (2) identification of information to be relinquished at the publish phase for the service, as either of these two tasks may support one or both of the complementary views to service composition pointed out above. Summarizing, SWS processes can be specified as precondition-effect pairs (supporting a planning approach to composition), or as abstract, observable communication behaviors, i.e. message type exchange sequences (supporting a choreography type of approach to service composition). In this research, a top ontology for SWSs will be developed in OWL with a Horn Description Logic base (OWL-DL with SWRL), supporting both types of SWS specifications. Translations of the service specification language will be developed in suitable process calculi (e.g. value passing CCS, or the π-calculus, as proposed in [8]). The semantic modeling of the specification language will be investigated (through a semantics appropriate for the representation process languages), in a suitable class of open petri nets. The problem of automating service composition will be revisited, using appropriate logical calculi (e.g. current developments on combinations of Propositional Dynamic Logic with the Logic of Information Flow [6]).