Semantic Interoperability

Interoperability is the ability of two or more systems or components to exchange infor-mation and to use the inforinfor-mation that has been exchanged [119]. Interoperability is a broadly used term, encompassing many of the issues impinging upon the effectiveness with which diverse information resources might fruitfully co-exists. The issues can be defined for different purpose, such as, semantics.

Semantic interoperability is the ability of two or more computer systems to exchange information and have the meaning of that information accurately and automatically interpreted by the receiving system. The main obstacle of semantic interoperability is semantic heterogeneity of the information to be exchanged. Common understanding of semantics and standardization of semantic representation are usually concerned as the solutions tackling the semantic heterogeneity to achieve semantic interoperability.

2.4.1 Semantic heterogeneity

Semantic heterogeneity is usually distinguished from syntactic heterogeneity and struc-tural heterogeneity in the database community [37] [39] [86] [88] [89]. Syntactic hetero-geneity is concerned with the heterohetero-geneity of data formats. Standardizing data formats is taken as an approach to solve syntactic heterogeneity problems. For example, XML is used as a standard format for all forms of Web-accessible data. Structural heterogeneity is associated with different data models, data structures or schemas, e.g. relational and object-oriented database models. An example of the solutions for structural heterogene-ity is that RDF based on XML syntax provides a unified way to structure information sources or object models for Web-based information exchange [100]. When two informa-tion sources are modeled in a same format by applying a same modeling methodology, there still might be semantic heterogeneity problem. Semantic heterogeneity can be identified according to the different types of conflicts [172]:

1. Semantic conflicts. Different modelers do not perceive exactly the same set of real world objects, but instead they visualize overlapping sets (included or inter-secting sets). For example, a "Student" object class may appear in one schema, while a more restrictive "CS-Student" object class (grouping students majoring in computer science) is in another schema. The "CS-Student" class will be integrated as a subclass of the "Student" class in the integration of two schemas.

2. Descriptive conflicts. Descriptive conflicts include naming conflicts due to homonyms and synonyms [7] [111], attribute domain, scale, constraints, operations, etc. [84].

5FOL (Fist Order Logic) can formalize the set theory.

2.4. SEMANTIC INTEROPERABILITY 23 3. Structural conflicts. Such structural conflicts are different from structural het-erogeneity. Even if two modelers use the same data model, they can choose different constructs to represent common real-world objects. For instance, in object-oriented models when a modeler describes a component of an object type O, he has the modeling choices between creating a new object type or adding an attribute to O.

2.4.2 Semantic annotation

The goal of empowering computer systems with semantic interoperability rests on the desirability of computer systems being able to find information and to use it for purposes that the original creator of the information did not anticipate. This goal of flexible information reuse requires some degree of understanding of the information, which in turn requires that the information be encoded in some standard fashion that is interpreted identically by all systems using that information. As a shared model of what the information represent, ontologies are usually used to achieve the level of understanding. Semantic annotation is an approach to link ontologies to the original information sources.

Annotation is the extra information associated with a particular point in a docu-ment or other piece of information. For semantic annotation, the extra information is meaning definitions of the concepts used in a document. The meaning definitions are in most cases represented in ontologies. Annotation can be in the form of comments, or in the form of metadata. Metadata is data about data and it is used to facilitate the understanding, use and management of data. Machine-manipulable annotations are often organized as metadata, which is also the format of semantic annotations. There are a number of alternative approaches regarding the organization, structuring, and preservation of annotations. For instance, all the markup languages (HTML, SGML, XML, etc.) can be considered schemas for embedded or in-line annotation. On the contrary, open hypermedia systems use stand-off annotation models where annotations are kept detached, i.e. non-embedded in the content. Both annotation approaches can be document-level (annotating the document as a whole) or character-level (referring just a specific part of the text) [81] (see Figure 2.4). Embedded annotation seems easier to maintain. However, non-embedded annotations allow dynamic, user-specific semantic annotations because they can change corresponding to the interest of the user or the context of usage. The embedded annotations might also have negative impact on the volume of the content and complicate the maintenance [70].

Figure 2.4: Embedded annotation and stand-off annotation [81]

24 CHAPTER 2. PROBLEM SETTING In [70], semantic annotation is used to establish links from the entities in the text to their semantic descriptions so that a number of basic prerequisite for representation of semantic annotations are identified:

• Ontology (or at least taxonomy) defining the entity classes. It should be possible to refer to those classes;

• Entity identifiers which allow those to be distinguished and linked to their se-mantic descriptions;

• Knowledge base with entity descriptions.

Semantic annotation of Web services has emerged under the hypothesis that se-mantics can improve software reuse and discovery, significantly facilitate composition of Web services and enable integrating legacy applications as part of business process integration [203]. Semantic annotation of Web services is also called semantic markup of Web services, for which numbers of semantic markup languages and approaches are proposed such as WSMO [209], METEOR-S [187], OWL-S [198], SWSA/SWSL [181], WSDL-S [203]. They can be categorized into: a) annotating information in WSDL with ontologies (METEOR-S, WSDL-S); b) formalizing ontologies of Web service as a Semantic Web services representation language (WSMO, OWL-S and SWSA/SWSL).

In [206], semantic annotation of process models is concerned as a prerequisite of the vision of Semantic Business Process Management, which is very close to our proposal.

It will enable (or enhance) additional functionalities, namely the discovery and auto-completion of process fragments, which lead to more effective modeling with respect to the reuse of existing process artifacts at the conceptual level. The executable process models can be partly generated from the conceptual business process models, which in-dicates there are underlying links between business process models and executable Web services. Semantic annotation of business process models could therefore enable more automation in the implementation phase because the corresponding Semantic Web ser-vices can be discovered automatically [206]. Although the work has just initiated and it is still an ongoing project, it shares the same vision with ours, i.e. semantic annotation can be also concerned as an alternative approach to achieve the semantic interoper-ability of semi-structured sources such as business process models, in spite of semantic annotations of unstructured sources (e.g. textual documents) and structured sources (e.g. WSDL described Web services). Efforts on the semantic enrichment of enterprise models by semantic annotations are also put by TG4 (Task Group 4: Semantic Enrich-ment of Enterprise Modeling, Architectures and Platforms) in EU project INTEROP (Interopreability Research for Networked Enterprise Applications and Software, FP6 508011) [62], in which the main achievable targets are the semantic interoperability for model exchange, model transformation and model traceability. As the contemporary work, our research shares some similar objectives and available technologies. Since we participated in the INTEROP project, our contributions are also devoted as part of the results in the project.