We implemented some sample Web Services for the processors (such as align_warp, reslice, slice, softmean and convert) rather than using beanshell scripts in the workflow. We also demostrated three different approaches to create the Second Provenance Challenge workflow in Taverna for best practice.
Our metadata and provenance is managed through a NG4J named graph, and we have used TriQL queries to complete the first provenance challenge. In order to meet the requirement of phrase one, our provenance data were collected in the XML notation and are provided as RDF (Resource Description Framework) files. Figure1 graphically presents the schema (also called the data model). This schema also shared under http://www.cs.man.ac.uk/~yuq/SecondChallenge/ProvenanceOntology/.
Figure 1 Provenance ontology abstract view in Protégé
The provenance for the three parts of the primary workflow (which are also recorded with required annotations for query 8 and query 9) are as follows:
In order to produce the complete provenance for part2 and part3, extra constant processors (such as ReslicedImages, ReslicedHeaders, AtlasImage and AtlasHeader) were used to carry the results from one workflow to another. For this reason, each part of the provenance RDF contains:
Figure 2 - Part 1 of the Primary Workflow
Figure 3 - Part 2 of the Primary Workflow
Figure 4 - Part 3 of the Primary Workflow
The provenance for the three parts of the secondary workflow (which has been modified as per for query 7) are as follows:
Compared with the provenance of the primary workflow runs, these documents contain the same processors except each convert processor has been replaced with two new processors, pgmtoppm and then pnmtojpeg, in order to convert from PGM to JPEG in workflow Part 3. The difference is documented in the adaptation specified by Query 7 and on our first challenge results page which associated with Provenance Query 7.
to top