Technology Review – OwlSight

Most of us are familiar with the Protege-OWL editor as a widely-used visualization tool for ontologies.  Recently, a new tool called OwlSight was released from the consulting firm of Clark & Parsia , a small cadre of engineers and researchers with ties to the Universities of Maryland and Manchester.

OwlSight is a web-based ontology browser implemented using GWT, Google's Web Toolkit, and incorporates Pellet, an open source reasoner also supported by Clark & Parsia. Their website describes OwlSight as a lightweight tool that serves as a sample of forthcoming functionality for Pellet.

Before diving into OwlSight, Google's Web Toolkit deserves a short discussion. GWT gives OwlSight some nice features right out of the box. First, there is no software to download. It is possible for anyone to launch OwlSight via a web browser without any installation work or dependency ramifications (although there is an interesting note about the need to warn Safari users for some reason). Second, OwlSight isn't tied to any operating system. Third, the GWT framework has the development resources of Google's engineering team behind it, a point that gives the tool additional credibility.

I immediately noticed how easy OwlSight was to use. Since it is limited to visualization, the simple design provides great value. The menu items are based primarily around loading either a sample ontology (OwlSight provides four of them) or one of your own (based on either a resolvable URL or the ontology text in full). The typical OWL serializations are supported: RDF/XML, N3 and N-Triples. OwlSight wasn't able to load several N3 or N-Triple samples that I tried, such as owl-rulesGenesis or the RDF test cases from W3C. RDF/XML examples worked without a hitch. Figure 1 provides a view of the tool displaying some high-level metrics, such as property type counts and DL expressivity. The left-hand pane shows both class and property tree views, displaying an ontology's subclass and subproperty taxonomies akin to Windows Explorer. The right-hand pane details any particular element, such as a class, and this is where the tool begins to shine.

As a web application, everything is hyperlinked. It is so simple to traverse an ontology by selecting classes, properties and individuals of interest as you work your way through the model. Also, it's easy to discern asserted and inferred statements because OwlSight uses various colored references on an uncluttered white canvas. My first reaction to these inferred statements was wondering how the reasoning engine determined them. OwlSight provides a ‘Why?' button that details the various logical paths traversed by the reasoner - very slick indeed.

From a business point of view, it's easy to imagine directing a client (or his engineering team) to OwlSight and the URI of an ontology and giving him the freedom to inspect the ontology whenever and wherever he wishes. It's not necessary to read user documentation as navigation is intuitive. I imagine using OwlSight to display ontologies to potential clients without being forced to delve into the details of the Semantic Web, ontologies and the like. I can readily present my OWL constructs as a rich data model that capitalizes on re-use, machine-accessibility and data integration.

There are several features that would be helpful in OwlSight. Using the browser's back and forward buttons to return to previously-viewed classes and properties would be beneficial, especially with corresponding hotkeys. When exploring inferred statements, OwlSight can only display up to five different reasoning permutations. I'd prefer to see all possible paths or at least be able to set a threshold setting.

OwlSight is a wonderful tool in its simplicity and web-friendly nature. In addition, you get the benefits of a reasoner without any additional work. You can try it out at: