Using Mapping Relations to Semi Automatically Compose Web ...

Using Mapping Relations to Semi Automatically Compose Web Services


The composition of web services is achieved by creating a third web service (hereafter
referred to as the integrating web service) and its Web Service Description Language
(WSDL) file. This web service invokes the source legacy web service to retrieve dat
a,
mediates to resolve any mismatches the data has with the destination web service,
typically through invocation of common services, and then passes the converted data to
this destination web service. The mediation step can take various forms including a

XSL
transformation. Typically the XSL is generated at design time by either a programmer or
by a data expert using an editor that permits the matching of entities in different schemas
by drawing lines between schemas entities. In data management tools s
uch as Clio, and
MetaMatrix, the matched entities correspond to logical constraints on the schemas,
making them difficult to exchange across systems.
Semantic mediation
and Semantic
Web Service approaches require

the use of ontologies to achieve

mediation

in
semantic
composition of Web services.
In this presentation, we demonstrate the generation of
mediation code or of the XSL through the use of mapping relations that are applied on
XML schemas. We use six mapping relations to align entities be
tween the

schemas. The
mapping relations are:
equivalentClass, sameAs, subclassOf, hasMatch, hasContext, and
hasRelation.

The
hasMatch

and
hasContext

relations are needed in order to resolve structural and
representational mismatches between the legacy schemas.
The
hasMatch

and sameAs are
used to reconcile structural and syntactic mismatches between legacy systems. The
hasRelation

establishes a generic relationship between a subject and an object.


Further, the embedding of the mapping relations in HTML

permits

the creation of
JavaScript libraries that abstract the invocation of web services, the handling of their
returned documents, and mediation between web services.

This gives birth to a new
breed of w
eb

pages

that combine

metadata
,

in the form of mapping re
lations
,

and
scripting.

In general the

Web page

con
tains four types of blocks:

1.

Imports of Web Service Description
Language (WSDL) files
, schemas,
ontologies, and other
mapping relations

2.

Alignments of entities and concepts

3.

Workflow statements

4.

Mediation statement



In addition to the benefits of mapping relations for

abstracting coding complexities, we
articulate the following benefits:


1.

T
he mapping relations are used to generate the schema or WSDL file of the
integrating Web service.
Today
this is done manually by programmers or Data
modelers.

2.

T
he mapping relations can be exchanged between different platforms simply by
copying them.

3.

Consequently, mapping relations enable reuse in service oriented architecture.

4.

The mapping relatio
ns enable XML Schemas as a

starting point for s
pecifying
semantics
. This is in contrast to solutions requiring ontologies as the starting
point for achieving data mediation.

5.

The mapping relations promote crosswalks between schemas for the purpose of
bu
ilding aligned ontologies.


6.

The mapping relations create a web of mashups. In contrast to the islands of
mashups that are emerging today, the propose
d approach gives birth to a web of
structured data, effectively addressing the deep Web problem. That is, the
boundaries that exist today between mashups created by different hosting
providers disappear with

this approach. Furthermore, Mashup editors can

be
supplemented with automated mediation capabilities.