The Information Workbench -- A Platform for Linked Data Applications

Tracking #: 539-1742

Authors: 
Anna Gossen1
Peter Haase
Christian Hütter
Michael Meier
Andriy Nikolov
Christoph Pinkel
Michael Schmidt
Andreas Schwarte
Johannes Trame

Responsible editor: 
Guest editors Semantic Web Interfaces

Submission type: 
Tool/System Report
Abstract: 
In this system paper, we describe the Information Workbench, a platform for developing Linked Data applications. The platform features a highly customizable user interface to present the data to the users and realize different interaction mechanisms. UI development is based on Semantic Wiki technologies, enriched with a large set of widgets for data access, navigation, exploration, visualization, data authoring, analytics, as well as data mashups with external data sources. Widgets can be easily embedded into Semantic Wiki pages in a fully declarative way using a simple wiki-based syntax. In this way, the structure and behavior of the user interface can be easily customized to create domain- and application-specific solutions with little effort. Our paper describes the technical architecture of the Information Workbench platform, particularly focusing on the tight coupling between the domain-specific semantic data model and the elements of the user interface, which in turn facilitates UI customization. Furthermore, we describe examples of applying the Information Workbench platform to build applications with semantic interfaces for real-world use case scenarios in different domains, including data center management and semantic content publishing.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Minor Revision

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Timo Stegemann submitted on 08/Oct/2013
Suggestion:
Minor Revision
Review Comment:

This paper describes a Semantic Web platform called "Information Workbench". In comparison to the previous version, this redacted version contains a "Lessons Learned" chapter as suggested in a preceding review. Here, the authors discuss technical limitations of the system. This analysis is a bit too shallow. The first limitation (reliability and serviceability) is very obvious to everyone who worked with public SPARQL endpoints. Furthermore, it reads like the authors already gave up solving the second limitation (scalability) by themselves.

Figure 1 still has some commercial feel. I would suggest changing at least the point "Your Solution" into something like "custom solution". Also, Figure 1 is referenced on page 3 to support the description of the SDK's components. However, the components are described in a completely different order than shown in the illustration. This is a quite confusing, especially because of sentences like "First, there are…" and "Second, the SDK provides…".

On page 5, the authors cite Keim's visual analytics mantra/paradigm: "initial analysis and overview first, details on demand". This quotation is not correct! The correct visual analytics mantra and therefore correct quotation is "Analyse First - Show the Important - Zoom, Filter and Analyse Further - Details on Demand" ("Overview first" is part of Shneiderman's information seeking mantra). This has to be fixed.

Apart from these critical remarks, this paper offers a good and clear description of the platform. Both paper and platform are relevant to the Semantic Web Journal readership.

Review #2
By Lloyd Rutledge submitted on 02/Dec/2013
Suggestion:
Major Revision
Review Comment:

For sake of completeness, I copy here my review of the previous version of this submission.

> Suggested Decision: Major revision
> Review Comment:
>
> This paper fits the category “Reports on tools and systems” described at http://www.semantic-web-journal.net/reviewers. It is a short paper, at just over seven papers, describing an implemented system. It is thus not a research paper; this paper offers no new research insights, and has no value other than the system it describes. In addition, the system itself is not a research contribution, but a system that applies many previous external research contributions, albeit hopefully in a novel way that provides an important impact.
>
> The tool is publically available for free-of-charge download and installation, but one must pay for additional features. This paper and tool’s website and online videos and demo show the system to be mature and of high quality. The paper is a clear and readable description of the tool. The system components and features reflect good knowledge of research developments in the field. The system described is one a person with good academic understanding of open data would want.
>
> Next to these positive aspects of this system report are critical remarks. For a publication in a research context, this paper has a relatively high commercial feel. The illustrations, particularly figures one and two, provide little academic information. Instead, they seem like slides from a presentation providing a colorful overview for the product.
>
> In addition, while the paper motivates the system’s components well in terms of start-of-the-art functionality as described by recent academic literature in the field, it lacks any critical perspective of its own system. There are no described limitations of the system appear, nor are there lessons learned, nor things the builders would have done otherwise looking back. The closest to these is some description of future work. A system paper in this journal should not try to convince every reader should download or buy its product, as this paper does. Instead, papers like this should provide information from which the reader can make an unbiased and informed choice about whether to apply this system for a given situation or another.
>
> A final limitation of this paper is the lack of convincingly described impact and importance. The paper does indeed present many features and components and describes adequately why such a system would want them. However, this paper does not convince the reader that its system as a whole has importance as a novel impact. It is thus not clear how the introduction of this system to the collection of similar systems changes the world of Semantic Web systems in a new way.
>
> On page four, “emedded” should be “embedded”. I have noticed no additional spelling or grammatical errors. However, I did not read the paper as an editor. The authors should take extra care to ensure there are no errors before publishing this paper anywhere.
>
> Comment for editor:
>
> Submitted on:
> 08/Jul/2013

The changes since then and this latest version of the submission are scattered typo-level changes, 7 scattered paragraphs with more details about features, and a new section with four paragraphs called “Lessons Learned”. This is not the amount or nature of changes I had in mind with my previous review. The new Lessons Learned section describes two shortcomings that are true of any LOD system. It then wraps up with a top-level or market-level description of its features. Thus, my original review still holds for this new version of the submission, except that I earlier said there are no “lessons learned”, and now there is a section with that name.

Such top- and market-level descriptions form the core of how this document falls short of a paper in a research journal throughout the text. This looks and runs like a nice product. It seems worthy to apply for setting op website interfaces to LOD. In addition, this document could serve this worthy product well as a good brochure motivating potential users to install and use it. However, this paper has negligible scientific value.

The feature description is very broad and high-level. It describes a full Semantic Web system from data to human interface. Of course, it is important that such systems exist, especially in the market. However, scientific interest usually lies in the details of components of a complete system, which this submission does not supply.

The phrase “semantic wiki” appears frequently, intermingled occasionally with “Semantic MediaWiki”. I assume that all semantic wiki features in the system Information Workbench as a whole come from use Semantic MediaWiki, which the Workbench builds on top of, probably as an extension in PHP. The document should provide this and similar details. What brings doubt to the novelty of the document’s system is that many of the types of interfaces and views the document describes are already implemented in Semantic MediaWiki and its extensions, especially Semantic MediaWiki Formats, but also many other extensions. It seems most if not all of the interface could be built with previously existing Semantic MediaWiki extensions. The paper should be clear about exactly what this system introduces as novel functionality, and exactly where the seam is between what this code introduces and what it loans from the externally acquired foundational software.

Here is an excerpt from the publicly available page on http://www.semantic-web-journal.net/reviewers and its section about “Reports on tools and systems”, for which this paper seems to qualify: “These submissions will be reviewed along the following dimensions: (1) Quality, importance, and impact of the described tool or system (convincing evidence must be provided). (2) Clarity, illustration, and readability of the describing paper, which shall convey to the reader both the capabilities and the limitations of the tool”. The quality of the system seems high, but both the importance and impact seem low because the novelty is unclear. Every feature in this system is recognizable as a feature in some other feature. Perhaps this system has value as a unique combination of these features, or a better industry-ready quality implementation of them. However, it would be more scientifically relevant if there were a new feature, or an existing feature implemented in a new and measurably improved way. At least, as a tool report, there should be something from the experience of using the tool that would help readers develop even better tools, but here that is lacking as well. The clarity, illustration and readability of this submission are good. However, again, they are good as a market-level description of a product.