The new W3C standard R2RML3 de nes a language for expressing mappings from relational databases to RDF, allowing applications built on top of the W3C Semantic Technology stack to seamlessly integrate relational data. A major obstacle in using R2RML, though, is the creation and maintenance of mappings. In this demo, we present a novel R2RML mapping editor which provides a user interface to create and edit mappings interactively even for non-experts.
the {gid} column, and specify mo:recording as the type of these subjects. The predicate rr:predicateObjectMap points to a structure de ning propertyvalue pairs for the subjects de ned by the rr:subjectMap; in the example, for every subject we create a property mo:duration, pointing to a literal created by column length of the source database.
Example 1. rr:logicalTable [rr:sqlQuery """SELECT id, gid, length FROM musicbrainz.recording
WHERE musicbrainz.recording.length IS NOT NULL"""] rr:subjectMap [rr:class mo:recording ; rr:template "http://musicbrainz.org/artist/{gid}\#\_"] rr:predicateObjectMap [rr:objectMap [ rr:column "length" ;
rr:datatype <http://www.w3.org/2001/XMLSchema\#float> ] ; rr:predicate mo:duration].
As can be seen from the example, the syntax to denote the mapping is rather convoluted and requires deep technical knowledge of R2RML, its language constructs, syntactic details (such as quotation), and the R2RML vocabulary itself. 3
The editor that we have developed in order to overcome these technicalities features an intuitive user interface which hides the R2RML vocabulary details from the user, provides access to relational metadata of the relational database for which the mappings are to be created, provides instant feedback where, at each step, the user has the option to preview the triples that would result from the mapping that is being created, and supports most of R2RML vocabulary except rr:graph, which we plan to include as we continuously evolve the editor. In the following we brie y describe the various steps of the editor's wizard work ow. Step 1 Datasource, Base URI Selection. The rst step in the wizard lets the user choose the datasource for which mappings shall be de ned, i.e. by design the editor supports the maintenance of mappings over multiple relational data sources. As part of the datasource selection, the user may also choose a base URI for URI construction templates speci ed in the upcoming steps.
Step 2 R2RML Rule Selection. After datasource selection, the editor loads and displays previously saved mappings for the datasource. At this point the user may choose to edit an existing rule or add a new one. The editor also provides a lter using which the mappings could be looked up based on involved table names, concept names and property names.
Step 3 Logical Table Selection. In this step, the editor provides various options to choose a logical table. The logical table could be a database table, view or a custom SQL query written by the user. If the option to add a database table is selected, all the tables from the database schema are displayed. Options to preview the table's data and metadata are also available. On the other hand, if the option of SQL query is chosen then an SQL editor is shown where the user can type SQL queries and preview the results.
Step 4 Subject Map Creation. Once a logical table has been selected, the editor guides the user to a subject template creation step. By the click of a button, the user can carry over columns that should be used to generate the template. Additionally, classes (i.e., types) can be added to the subject by using the auto-complete eld rdf:type, which displays all the classes that are present in the system. The editor is exible enough such that multiple classes can be added. At this stage a preview of the generated triples can be accessed as well. This provides feedback to the user such that they can adjust the values in order to obtain the intended results.
Step 5 Predicate-Object Maps Creation. After the subject map creation, predicate-object maps can be added to the mapping. Again, the editor provides exibility for adding as many predicate-object pairs as needed. Additionally, for object maps all types of terms de ned in R2RML are supported, namely constant terms, column values, templates, and object references. The object reference option lets the user choose from the previously de ned mappings. Furthermore, on choosing the appropriate rule the child and parent columns are prepopulated from the corresponding logical tables. Again, multiple join conditions can be added through the interface. The preview triples option is also available at this stage, again providing the user with valuable feedback. Fig. 1 depicts the screen corresponding to this step.
Step 6 Textual Representation. Finally, an intuitive textual representation of the mapping rule is displayed. At this point the user could go back to any previous step to modify the mapping or otherwise save the mapping.
The demo video: https://www.youtube.com/watch?v=PfP2wmWw9-o is available online, the video includes a very brief introduction to R2RML and motivation for an editor to create R2RML mappings. The remainder of the demonstration shows a use case of mapping MusicBrainz5 database tables to the Music ontology.6 In particular, we demonstrate two example mappings, where each step of the mapping creation process is explained and special features of the editor are highlighted.
Although tools like Karma [2], Ontop [3], Topbraid Composer7, and Neon
Toolkit8 can be used for mapping relational and other datasources to RDF,
the editor presented here provides the following advantages (