MOSIS : Modelling Semantically Interoperable Web Service Compositions

The use of information from different, often distributed, sources is part of many applications today, and in many cases these are heterogeneous sources developed independently without the intention of serving one specific application. This introduces the problem of semantic heterogeneity. Different sources are usually based on different vocabularies even in cases where they operate within the same domain. Data items defined in one source may have different definitions in another source although they really are the same (synonyms), and items that are different in reality may have similar definitions in two different sources (homonyms). There is a need for using these sources with regards to the meaning of the items in question not how they are defined in the sources, thus achieving what is called semantic interoperability. The distributed sources that will be treated in this thesis are Web Services. Web Services provide a programmatic access point through the web to a set of related functionalities, and these can be combined to create a new Web Service with added value. The combination of Web Services is called a Web Service composition. Web Service composition impose several semantic interoperability issues. First of all Web Services need to be discovered, you can use it if you don't know it exists, and this requires a semantical match between the descriptions of a requested service and the descriptions of existing services. Secondly, when you compose service, you need to make sure that information from one service that will be used as input to another service is semantically related to the input. In order to deal with semantic interoperability issues concerning Web Service compositions we propose MOSIS (MOdelling Semantically Interoperable web Service compositions). MOSIS makes use of UML models and principles from Model Driven Architecture (MDA) to model the discovery, the composition and the ontologies used to define semantically the concepts involved. As an effect of using UML modelling we claim that both the process of using ontologies and defining them will become easier for developers unfamiliar with ontology description and developing. MOSIS facilitates functionality based Web Service discovery which makes it possible to match descriptions of a Web Service with descriptions of existing Web Services. In addition to this MOSIS provides measures of dealing with mappings that have to be developed in order to make sure that output from one service can be used as input to another