Home

Documenting Scientific Processes with Scientific Workflows

It is important for scientists to document the scientific processes that they use to generate scientific artifacts. It is also important for other scientists to understand the processes that were used to produce scientific artifacts. Scientific workflows is one approach to document scientific processes. Many workflow management systems require scientists to document their scientific processes at a level of detail that requires technical prowess, which may be distracting from the main goal of capturing scientific knowledge.

The approach of this work is to create task ontologies to capture domain knowledge that effectively represents a controlled vocabulary of a project, as well as additional knowledge that suggests the use of this vocabulary towards the description of processes for that project. Next, scientists use the knowledge encoded in the ontology to document scientific processes in the form of abstract workflows. The ontologies used in this approach are referred to as Workflow-Driven Ontologies (WDOs), and the WDO-based workflow specifications are referred to as Semantic Abstract Workflows (SAWs).

Workflow-Driven Ontologies (WDOs)

WDOs are task ontologies encoded in OWL. Their intended use is to document concepts about a domain for the purposes of capturing process knowledge. The two main classes of WDOs are Data and Method. The Data class is representative of the data components of a scientific process. Examples of data components are datasets, documents, instrument readings, input parameters, maps, and graphs. The Method class is representative of discrete activities involved in a scientific process to transform data components. Examples of discrete activities are algorithms (either implemented as software or carried out manually by humans) and human inspection of data. Scientists document process-related concepts by extending the hierarchies of Data and Method and by defining isInputTo and isOutputOf relations among the process-related concepts to capture their data-flow interdependencies with respect to a scientific process.

As an illustrative example, the figure below shows the WDO concept hierarchies for the process of creating a uniformly-distributed geospatial dataset from a sparsely-distributed dataset.

The OWL encoding of this WDO can be found at http://trust.utep.edu/2009/gridding/wdo and the WDO-It! tool provides a GUI to view and edit it.

Semantic Abstract Workflows (SAWs)

Scientific workflows can be generalized as graphical structures that contain nodes representing discrete activities and directed edges representing data flow between those activities. Traversing the graph from its initial data sources to its final data sinks simulates the action of carrying out a complex process conformed of simpler activities.

SAWs are the proposed artifacts for scientists to capture scientific processes. Semantic refers to the meaning inherited by using ontological concepts captured in a WDO. Abstract refers to the fact that the workflows captured lack the additional constructs necessary to produce automated systems. In this sense, SAWs are not committed to be executable workflow specifications.

As an illustrative example, the figure below shows a SAW that uses the concepts presented in the figure above to describe the process of creating a uniformly-distributed geospatial dataset from a sparsely-distributed dataset.

Notice that the labels have the format XXX:YYY, where XXX corresponds to the name of the concept defined in the WDO, and YYY corresponds to an assigned name for that particular instance of the concept. Furthermore, the ovals represent data Sources and Sinks, and their types (i.e., Dataset and Person) are not defined in a WDO, but are concepts reused from the PML-P ontology, a provenance ontology from the Inference Web project.

The OWL encoding of this SAW can be found at http://trust.utep.edu/2009/gridding/gridDataset and the WDO-It! tool provides a GUI to view and edit it.

Please see the documentation or contact us for additional details.