Introduction - User-Level Workflow Design: A Bioinformatics Perspective - page 11

This topic is concerned with the simplification of workflow design in the
upwards direction, that is, with bringing workflow design closer to the user.
The central challenge that has to be addressed in this respect is how to
“teach” (domain-independent) workflow management systems to “speak the
language of the user”. In fact, “many scientists would prefer to focus on
their scientific research and not on issues related to the software and plat-
forms required to perform it” [193]. However, the level of abstraction that is
common to the present systems still requires quite some technical knowledge
about IT and programming in general and the services in particular. Mark
Wilkinson's statement that “Many biologists become bioinformaticians out
of necessity, not because they like computer science.” reflects that the current
user group consists of IT-ane application experts and computer scientists
rather than of the end-users themselves. Enabling the latter to effectively
work with analysis workflows on their own means to leverage the workflow
design to an even higher, more conceptual level which truly abstracts from
the underlying technologies. That is, application-specific terminology rather
than computer science vocabulary should be used to annotate data types and
services as described in the previous sections, so that users are able to work
with a world-wide distributed collection of tools and data using their own
domain language.
As indicated in Figure 1.5 (right) and detailed in the following, this challenge
can be addressed by working towards the semantic handling of services, data
and workflows. Semantic handling of services provides the basis for user-level
workflow design by facilitating the discovery of services that are adequate for a
particular situation. Semantic handling of data goes further and considers the
relationship between services in order to reason about their compatibility .The
“supreme discipline”, however, is their combination with the user's intents ,ex-
pressed via constraints in terms of his domain language, to achieve semantic
handling of entire workflows. This allows for assessing the adequacy and cor-
rectness of existing workflows and, even more, enables the automatic creation
of new workflows that are correct and adequate by design.
All these approaches have in common that they depend on the availability
of proper semantic characterizations of the resources in the application do-
main. That is, a domain model [243] is required that provides a formal concep-
tual model of the application domain. Typically, a domain model comprises
a domain-specific vocabulary (usually in the form of an ontology), resources,
resource descriptions in terms of the domain vocabulary, a definition of the
(possible) relationships between the resources and any additional constraints
that characterize the application domain further. In the context of domain
modeling for workflow applications, the resources are simply the services and
data of the application domain, and the concept of how to relate resources
is provided by the workflow management environment, which already deter-
mines how services and data can be assembled into workflows. Accordingly,
domain modeling in this work focuses on service integration, domain-specific
vocabularies, resource descriptions and constraints.