Automated Composition of Scientific Workflows: A Case Study on Geographic Data Manipulation

摘要

Scientific workflow management systems have become popular in recent years to support scientists in the assembly of computational pipelines [1]-[3]. Automated workflow composition [4]-[8] aims at adding even another level of abstraction that lets scientists compose purpose-specific workflows without having to find out which tools can be combined and how. Instead, they use abstract terms from a domain ontology to describe the intended workflow in a declarative way. Specialized composition algorithms translate the specifications into concrete and executable workflows. The successful application of these algorithms depends on the availability and quality of semantic domain models that capture knowledge about the tools and data in the domain. However, they are usually not readily available. The bioinformatics community made considerable progress in this context in the last years, with the EDAM ontology [9] providing a controlled terminology for describing operations and data types, and repositories like bio.tools [10], which list computational tools along with their semantic annotations. From there, automated workflow composition is straightforward, as for example demonstrated in [11]-[ 13]. To the best of our knowledge, no such ontologies and annotated tool repositories are yet available in the domain of geographic information and data processing.