LabelFlow: Exploiting Workflow Provenance to Surface Scientific Data Provenance

Pinar Alper (1) , Khalid Belhajjame (2) , Carol Goble (3) , Pinar Karagoz (4)

1 University of Manchester
2 LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision
3 CS - Computer Science Department [Manchester]
4 Universite Bilkent [Ankara]

Provenance traces captured by scientific workflows can be useful for designing, debugging and maintenance. However, our experience suggests that they are of limited use for reporting results, in part because traces do not comprise domain-specific annotations needed for explaining results, and the black-box nature of some workflow activities. We show that by basic mark-up of the data processing within activities and using a set of domain specific label generation functions, standard workflow provenance can be utilised as a platform for the labelling of data artefacts. These labels can in turn aid selection of data subsets and proxy for data descriptors for shared datasets.

Mots clés: Provenance; Annotation; Scientific workflows

Domaines: Informatique [cs]

Pinar Alper, Khalid Belhajjame, Carol Goble, Pinar Karagoz. LabelFlow: Exploiting Workflow Provenance to Surface Scientific Data Provenance. 5th International Provenance and Annotation Workshop, IPAW 2014, Jun 2014, Cologne, Germany. pp.84-96, ⟨10.1007/978-3-319-16462-5_7⟩. ⟨hal-01343407⟩