Pursuing the goal to lower the entry barrier into the world of Linked Data, we present the Information Workbench as a platform to support self-service Linked Data application development. Targeting the full life-cycle of Linked Data applications, it o ers support in discovery and exploration of Linked Data sources, facilitates the integration and processing of Linked Data following a Data-as-a-Service paradigm (where remote data sources can be virtually integrated through a federation layer), and eases self-service UI development based on Semantic Wiki technologies, combined with a large set of widgets for interacting with the data. Coming with all these features, the Information Workbench can be used to rapidly build industrial-strength Linked Data applications.
In recent years, a large amount of Linked Data (LOD) has been published on the Web [2]. Growing in size and domain coverage, this data becomes more and more interesting for building innovative applications that integrate heterogeneous data from di erent sources, in order to overcome the limitations of traditional data management systems. Apart from a new era of Web applications that exploit the LOD corpus, this development also o ers opportunities in building novel applications for the enterprise by bringing together company-internal data with external data, to augment and contextualize internal knowledge bases [3].
The development of speci c applications that bene t from Linked Data, though, comes with a variety of new challenges. First, at the data management side, developers are faced with a variety of new data formats and query languages (such as RDF, OWL, and SPARQL), but also struggle with heterogeneity at data level (facing Linked Data available via HTTP lookups, RDF dumps, and SPARQL endpoints) and new database systems and tools to store, process, and access this data. Second, once the relevant data has been identied and integrated into the system, Linked Data applications require new data interaction paradigms to deal with the speci c challenges { and opportunities { of the underlying data formats, such as schema exibility and data semantics. In particular, to leverage the bene ts of Linked Data, aspects such as dynamic discovery of data sources, seamless integration of Linked Data from multiple sources, provenance, application development environments, and { last but not least { end-user interfaces that implement generic interaction paradigms with Linked Data are important aspects when building Linked Data applications.
In this paper, we present the Information Workbench as a platform to support
self-service Linked Data application development. We start with a discussion of
the Information Workbench architecture in Section 2. In Section 3, we discuss
how the self-service idea empowers all stages of the application development
process, in particular (
{ For every data source, a data provider can be instantiated. Such providers gather information from the source, convert it into the RDF data format, and materialize the RDF output in the central triple store. Apart from built-in mechanisms to integrate semantic data formats such as RDF dumps or data behind SPARQL endpoints, there are several generic providers supporting the integration of legacy and Web data sources. { Data can be manually loaded in the platform using prede ned UIs and APIs for data import. Supporting the integration of tabular data and spreadsheets, the Information Workbench also o ers interfaces to Google Re ne. { Finally, our platform supports virtualized data integration, where local or public Linked Data sources (such as SPARQL endpoints) can be connected through a federation layer without materializing the data in the central store.
Once the data has been integrated, every resource in the data graph is automatically associated with a Semantic Wiki page, making it possible to bring together the structured data of the resources with unstructured information. Overcoming the limitations of traditional wikis, Semantic Wikis o er built-in mechanisms to access the underlying data, namely to extract and display information from the underlying data graph (e.g., by means of SPARQL queries) and to write directly to the store (e.g., by semantic links that connect resources at data level). Accounting for the coexistence of structured and unstructured data, the platform implements advanced search and information access paradigms, ranging from keyword search to complex graph pattern-based search, supporting the user in constructing expressive search queries by use of forms and state-ofthe-art semantic query auto-completion techniques.
All the core system functionality is extensible and exposed to the outside by di erent APIs, including Java interfaces, an integrated SPARQL endpoint, or an interactive CLI. The Information Workbench also comes with an extensible AJAX-based Web frontend, which supports mashups on both data and widget level, making it possible to interlink data from multiple sources using di erent visualization and exploration widgets. 3
The Information Workbench supports the whole Linked Data application
development process. This process is aligned with the self-service idea, thus hiding
the technical details behind data integration, data management, as well as the
complexity of UI building. In particular, the platform o ers support for (a)
selfservice data integration, (b) self-service analytics (comprising aspects like
userde ned, mashed-up interactive dashboards, e.g., in the form of charts, time-series
diagrams, geo-mappings, etc.), and (c) the ability to explore data and problems
collaboratively and interactively in real time. In the following we describe how
the Information Workbench supports self-service application development along
the whole application development process. We illustrate the process using an
example application in the media domain that intends to provide an end-user
oriented music portal built on top of di erent public Linked Data sources.
(
In our example application, we select a template designed for the media
domain, which is con gured to use the data from the LinkedBrainz project (a
Linked Data version of MusicBrainz.org data) as well as the DBpedia dataset.
The template-based provisioning of the Information Workbench (including the
local data repositories) is completed in about 5{10 minutes.
(
Building upon these prerequisites, Figure 2 shows how the Information Workbench implements the DaaS paradigm. At the bottom is the data layer, which includes public data sources from di erent publishers as well as local data sources (typically in the form of SPARQL endpoints). An integral component of the Information Workbench is a metadata registry, which provides a unied view on metadata and statistics about these data sets by integrating information from di erent data registries and markets such as http://ckan.org, http://data.gov, and others. As standard vocabulary for representing the metadata we use DCAT and VoID [1]. The user can explore the metadata catalog from within the Information Workbench along all meta information that is available, including domain and size of the data sets, origin, licenses, information about interlinked sources, etc. Figure 3(a) shows a screenshot of the PivotViewer interface for visual exploration of available data sources. In our example, we are interested in an additional dataset from the media domain: The BBC Music data set. Selecting the data set takes us to the detail page for BBC Music (cf. Figure 3(b)), with a description, statistical data, as well as information about the distribution of the data set. Once a relevant data source has been identi ed, its data can be integrated into the system by the click of a button. Depending on the access mechanisms that are supported for the data set, the Information Workbench o ers options to either load data into the local repository (if an RDF dump is available) or to connect the data virtually through a federation layer (whenever there exists an open SPARQL endpoint). The integration approach is transparent to the end user, which means that (i) within the deployment process the user does not need to be concerned with aspects of physical distribution, access protocols and interfaces, underlying data models etc., and (ii) the details of the integration are hidden at runtime, so both local data and virtually integrated data sources can be queried and accessed in an integrated way.
Virtualized data integration is realized by the use of FedX, a federation layer
for Linked Data, which we developed speci cally for the transparent access to
Linked Data [4]. FedX is a practical framework that incorporates novel
optimization techniques for e cient federated query processing in a distributed setting
and supports the ad-hoc integration of data sources at runtime.
(
The user interface can be customized using a rich pool of widgets that are shipped with the Information Workbench, targeting di erent data interaction paradigms such as semantic search, data visualization (e.g., as tabular results, graphs, charts, timelines, maps, heatmaps, etc.), navigation and exploration of the data (e.g., a graph-based data browser or a PivotViewer), collaborative editing, knowledge acquisition, as well as mashups with external data sources (e.g., Youtube, NY Times data, Facebook, and Twitter). All widgets can be easily embedded into Semantic Wiki pages and are speci ed using a declarative wiki-based syntax. With very little e ort, the standard views can be customized to create domain- and application speci c interfaces. Figure 4 shows a screenshot of the page of the Red Hot Chili Peppers. It is based on a wiki-based template de nition that describes how resources of type musical artist are presented. The chart at the bottom, for instance, is generated by the following widget declaration: {{#widget: Chart | query ='SELECT (COUNT(?release) AS ?count) ?label WHERE { ?? foaf:made ?release .
?release rdfs:label ?label . } GROUP BY ?label ORDER BY DESC(?count) LIMIT 30' | chart = 'BAR_VERTICAL' | input = 'label' | output = 'count' }}
In addition, the page includes a description of the artist taken from a remote Web service (Last.FM) and an embedded live video from YouTube for this artist.