Linked Data generation and publication remain challenging and complicated, in particular for data owners who are not Semantic Web experts or tech-savvy. The situation deteriorates when data from multiple heterogeneous sources, accessed via di erent interfaces, is integrated, and the Linked Data generation is a long-lasting activity repeated periodically, often adjusted and incrementally enriched with new data. Therefore, we propose the rmlworkbench, a graphical user interface to support data owners administrating their Linked Data generation and publication work ow. The rmlworkbench's underlying language is rml, since it allows to declaratively describe the complete Linked Data generation work ow. Thus, any Linked Data generation work ow speci ed by a user can be exported and reused by other tools interpreting rml.
Administrating the integration of the ever-increasing amounts of data from
multiple sources in di erent formats into a common knowledge domain remains
challenging and complicated, in particular for data owners who are not
Semantic Web experts or tech-savvy [
To minimize the e ort and knowledge that data owners need to administrate
their data and the overall Linked Data generation and publication work ow,
we developed a multi-user browser application, the rmlworkbench. This demo
shows how data owners can use the rmlworkbench. Depending on their assigned
roles, data owners can view and manage di erent sources for retrieving raw data
and the corresponding mappings to generate Linked Data, in contrast to our
earlier work on the rmleditor [
The FluidOps Information Workbench1, Ultrwrap2 and LinDa workbench3 are gui tools supporting data owners to generate Linked Data. However, the latter two only support tabular data, while the former, even though it supports more data sources, does not allow specifying di erent access interfaces. Most importantly though, none of them allows users to export their speci ed Linked Data generation work ow in a declarative, complete and interoperable way that allows to replicate the same Linked Data generation by other tools.
Linked Pipes4, and its predecessor Uni ed Views5, are general-purpose tools that allow users to administrate, execute, debug, monitor and share Linked Data processing tasks, for smooth and e cient management. However, they are not focused on Linked Data generation. They perform direct mappings which are afterwards processed via sparql construct queries. Moreover, they only allow to export the di erent processes descriptions, using their own custom descriptions.
The Silk Workbench6 follows a similar approach as the rmlworkbench. Even though it is a gui supporting users to administrate rdf dataset linking, it also requires corresponding aspects to be speci ed. Its function relies on projects which consist of linkage rules associated with data sources (data dumps or sparql endpoints) constituting altogether linkage tasks, as mapping rules are associated with data sources constituting generation tasks in the case of the rmlworkbench. 3
The rmlworkbench design principles are generic, following the classical
multitier client-server architecture. Its underlying language to declaratively de ne the
Linked Data generation work ow speci ed by the user is rml. rml [
The rmlworkbench consists of ve panels: Access, Retrieve, Generate, Publish and Schedule. In the remaining of this section, each panel is brie y presented. Access panel. Users can manage their own sources, which can be accessed through interfaces for local les, databases or Web sources. The descriptions are annotated using di erent vocabularies, e.g., dcat, csvw or Hydra. For instance, a user speci es a database accessed via a certain jdbc, and labels it \DB Source" and a dataset published on a dcat catalog, which he labels \Catalog Source". Retrieve panel. Not all data which appear in a data source are required to generate the desired Linked Data. Distinct subsets may be considered separately for generating di erent Linked Data sets. The Retrieve panel allows users to specify which exact data is retrieved for each selection. For instance, via the Retrieve panel, the aforementioned user speci es the exact tables, which are eventually considered to generate certain Linked Data. The user speci es the \Singers" and \Albums" tables of the \DB source" and labels them as \Singer data" and \Album data" respectively. Moreover, the same user speci es and labels as \Performance data", among the di erent datasets of the \Catalog source", the dataset about performances, and precisely its xml distribution.
Generate panel. To generate the desired Linked Data, the users need to specify sets of mapping rules. The rmlworkbench allows users to (i) upload a mapping document, (ii) specify a Web source with mapping rules, or (iii) directly edit them via its interface. Di erent sets of mapping rules may be associated with the same data, generating thus di erent Linked Data views. Once the set of mapping rules is associated with some raw data, the users can execute the mapping and generate the desired Linked Data (\Execute" button, as shown in the following gure). The dataset is then listed among the datasets available for publishing, or the users are noti ed if the generation was not successful. The users can specify mapping rules, for instance the sets of \Singer mappings" and \Performance mappings". Once the mapping rules are listed among the available sets, the users can associate them with the corresponding data (\Add Logical Source" button), in our example the \Singer data" and \Performance data". Furthermore, the users may desire to generate another Linked Data set with the same data. In that case, another set of mapping rules is added, e.g., the \Person mappings", and the user associates it with the \Singer data" as well.
Publish panel. A frequent activity after generating Linked Data is its publication. The rmlworkbench supports users to easily accomplish this activity. In our example, the Linked Data is published via an ldf server10. Nevertheless, the administrator can easily con gure other interfaces, for instance sparql endpoints. The users can then choose one or more of them to publish their Linked Data. Schedule panel. In most cases, the Linked Data generation and publication is a recurring activity. Data owners periodically regenerate their Linked Data set to keep it up-to-date with the original data. The rmlworkbench allows data owners to schedule the Linked Data generation and publication activities.
To summarize, the rmlworkbench allows data owners to specify their
complete Linked Data generation work ow, without being restricted by the tool. In
the future, the rmleditor [