J.-L. Ambite and D. Kapoor, Proceedings of ISWC 2007.
Presenters: Eran Chinthaka and Joseph Morwick
Presenters: Eran Chinthaka and Joseph Morwick
Presentation slides are available to download in ppt format.
This paper expands on previous work in workflow generation via semantic service descriptions (such as that by Gil et al) by considering use of relational descriptions and introduction of shim services. In the introduction, it is noted that previous work has either applied semantic classifications to the service as a whole or individual inputs and outputs, and that these approaches ignore the related nature of the inputs and outputs of services.
Their system uses a first-order theory to describe relational inputs and outputs for services within the bounds of a general ontology. This information is then used to automatically generate workflows for a given goal via a planning algorithm using PowerLoom as the inference engine. Shim-services are used to translate outputs of one service to meet input requirements of another. Many of these are "Domain-Independent Adaptors" meant for general application, such as the "Selection" shim which selects more specific data from a service which provides additional data unnecessary for for the service to-be-linked. Others are "Domain-Dependant" having a task specific only to the given domain. These are viewed as any other service by the planner. Lastly, "Generic Domain-Independant Adaptors" bridge the gap between the previous two. These adaptors can be applied in multiple domains but require additional knowledge to be useful. One example of this is unit conversion, where the task is generic but a specific conversion table is needed for any particular domain.
Finally, an evaluation of the computational efficiency is presented at the end of the paper. using hundreds of concepts and tens of relations and services. It took the system up to several minutes to plan workflows with up to 54 edges and 71 links.
Related Work
- Wings for Pegasus: Creating Large-Scale Scientific Applications Using Semantic Representations of Computational Workflows, Y. Gil, V. Ratnakar, E. Deelman, G. Mehta, J. Kim. Proceedings of IAAI 2007
- Transparent Grid Computing: A Knowledge-Based Approach Jim Blythe, Ewa Deelman, Yolanda Gil, Carl Kesselman Proceedings of IAAI-03
- Argos: A Framework for Automatically Generating Data Processing Workflows. Jos´e Luis Ambite Dipsy Kapoor
- Automatic generation of data processing workflows for transportation modeling. Ambite, J. L., and Kapoor, D. In Proceedings of the 8th Annual International Conference on Digital Government Research (dg.o2007), Philadelphia, PA, USA, 2007.
- Using ontologies for data mining workflow composition Rahul Ramachandran.
