Article Preview
TopIntroduction
The increasing availability of a broad variety of Web services (WS) raises the need to automatically discover, select and orchestrate appropriate services for a given need. Current WS search engines or service registries (e.g., UDDI) mainly support simple keyword-based search on Web services based on syntactic descriptions such as WSDL. However, this rather syntactic paradigm does not support precise allocation of Web services partially because of the lack of semantics expressed in utilized service descriptions. Semantic Web Services (SWS) (Fensel, Lausen, Polleres, de Bruijn, Stollberg, Roman & Domingue, 2006) aim at addressing this challenge on the basis of comprehensive, machine-interpretable semantic descriptions. Existing SWS frameworks, such as WSMO (WSMO Working Group, 2004) or OWL-S (Joint US/EU ad hoc Agent Markup Language Committee, 2004), enable the description of several WS-related functional (e.g. input and output, pre and post conditions, service choreography and orchestration) and non-functional (e.g. Quality of Service) parameters. However, since Web services usually are provided by distinct and independent parties, the actual WS interfaces as well as their semantic representations are highly heterogeneous. This strongly limits the interoperability and raises the need of mediating between semantic descriptions as well as the actual Web services interfaces. We can particularly identify two levels of mediation: semantic-level and data-level mediation. Whereas the former refers to the resolution of heterogeneities between concurrent semantic representations of services – e.g. by aligning distinct SWS representations of services equivalent in functionality – the latter refers to the mediation between mismatches related to the Web service implementations themselves, i.e. related to the structure, value or format of I/O messages. Therefore, semantic-level mediation primarily supports the discovery and selection stage, whereas data-level mediation occurs during service orchestration and invocation.
In this paper, we particularly address semantic-level mediation to support the WS selection problem. We argue that semantic-level mediation strongly relies on identifying semantic similarities between entities across different SWS ontologies (Qu, Hu & Cheng, 2006; Wu, Ranabahu, Gomadam, Sheth & Miller, 2007). However, semantic similarity is not an implicit notion within existing ontology representations. Moreover, automatic similarity detection as demanded by semantic mediation requires semantic meaningfulness. But the symbolic approach – i.e. describing symbols by using other symbols without a grounding in the real world – of established ontology representations does not fully entail semantic meaningfulness, since meaning requires both the definition of a terminology in terms of a logical structure (using symbols) and grounding of symbols to a conceptual level (Cregan, 2007; Harnad, 1999).
Despite the importance of mediation for widespread dissemination of SWS technologies, related approaches are still limited and underdeveloped (Paolucci, Srinivasan & Sycara, 2004). Current attempts to mediation usually foresee the manual development of rather ad-hoc one-to-one mappings or the application of ontology mapping methodologies, mostly based on identifying (a) linguistic commonalities and/or (b) structural similarities (Choi, Song & Han, 2006; Noy & Musen, 2003). Since manually or semi-automatically defining similarity relationships is costly, current approaches are thus not capable to support SWS selection within highly dynamic scenarios and at Web scale.