Introduction - User-Level Workflow Design: A Bioinformatics Perspective - page 16

a service that loads an svg image from a database, for instance, can not
(directly) be used in conjunction with the ImageMagick services, as they can
not process vector formats 1 .
Semantic Handling of Workflows: Intent
What is more, speaking the language of the user does not only involve the
component level, but also the whole workflow. Handling entire workflows in
terms of its semantics has two principal flavors:
1. Semantic information about the application domain can be applied for
monitoring workflow development , essentially by a kind of domain-specific
model checking [70], where workflow design is accompanied by the con-
tinuous evaluation of constraints that express domain-specific knowledge.
Similar to a spell checker in a text processing program, which hints at
the violation of orthographic rules, perhaps already indicating a proposal
for correction, the model checker notifies the workflow designer when the
defined constraints are violated. This provides truly helpful workflow de-
velopment guidance for the user, beyond mere validation of the workflow
model in terms of syntax checks.
2. Semantic information can be used for semantics-based, automatic
workflow composition : Rather than defining a workflow that already con-
stitutes the solution to a research question, users should only need to ex-
press the research question itself using domain-specific terminology, which
is then automatically translated into an executable workflow. Thus, not
only automation of the workflow would be achieved, but also of its de-
sign. Concretely, declarative workflow models as proposed, for instance,
in [325, 121, 246, 247, 257, 178, 225] can be used to capture the intents
about workflows at an user-accessible level. The concretization of the high-
level workflow specifications into executable workflow instances can then,
for instance, be realized by the application of synthesis-based [252, 163]
or planning-based [266, Chapter 10] composition approaches, which work
on domain-models that are derived from formally defined service and type
semantics and comprehensive (ontological) domain knowledge.
Figure 1.9 illustrates the notion of semantic handling of workflows, again
by means of a variant of the simple birthday card creation workflow. Here, the
user's intent can simply be expressed as “Take a picture from (his) camera and
send it as birthday card to Anna.”, as shown in the upper part of the figure.
Accordingly, the user may now simply connect an available service that loads
apicturefromacameratoaservicethatsendsapictureasabirthdaycard,
as shown in the center of the figure. Although this combination of services is
1 Principally, ImageMagick also supports vector formats, but this support is not
complete. Hence, the ImageMagick services defined for the example are simply
restricted to raster formats to avoid problems.