Article Preview
TopIntroduction
An important class of web services consists of those that primarily do data processing and produce output data (Saleh et al., 2009; Truong & Dustdar, 2009). This type of Web service is known as Data-as-a-Service. DaaS services return collections of Data for a given set of parameters without any side effects (Carey, 2006). DaaS services composition is a powerful mean to answer users' complex queries.
Several approaches are proposed to enable automatic Web service composition (Rao & Su, 2005). The Semantic-based ones proceed by describing the Web services properties over ontology. In fact, many ontology languages like OWL-S1 or WSMO2 and extension mechanisms like WSDL-S (Akkirajuet al., 2005) or SAWSDL3 provide standard means by which WSDL (Web Service Description Language) document can be related to semantic description.
However, these solutions do not provide a way to relate semantically the Web service parameters (i.e. input and output) which hampers their applicability to DaaS composition. The automation of DaaS composition requires the specification of the semantic relationships between inputs and outputs parameters in a declarative way.
This requirement can be achieved by describing DaaS as view over a Domain Ontology (DO) following the mediator-based approach (Wiederhold, 1992). In this context, several works (Barhamgi et al., 2010; Zhou et al., 2008; Vaculin et al., 2008) consider DaaS as Parameterized RDF View (PRV) with binding pattern over a DO. The PRV describes how the input parameters of the DaaS relate to the data it provides. Defined views are then used to annotate DaaSs description files (e.g. WSDL files) and are exploited to automatically compose DaaSs. Thereby, the DaaS composition problem is reduced to a query rewriting problem largely studied in data integration field (Halevy, 2001).
However, there are several references ontologies which formalize the same domain knowledge. Thus, the construction of a DO, unifying all existing representations of real-world entities in the domain, is a strong limitation to interoperability between DaaSs.
This limitation raises semantic conflicts between pieces of data exchanged during DaaS composition. To this end, the applicability of the previously cited DaaS composition approaches is not practical.
Therefore, considering the semantic conflict detection and resolution during the composition process is crucial.