MarBlog

In the last couple of posts I have been talking about ontology tools. In the meantime, I have been working a bit with Protege, getting a basic skeleton up (50 entitites/classes, similar number of instances, and a handful of object relations, and some data fields associated). Now, however, there is a point where I need to start sharing it with a group of colleagues. Not everyone in this group will be au fait with running protege and delving into the bowls of OWL files.

At first I looked at the simple exporter function, OWLDoc that will drop your ontology in plain HTML and which will probably end up being the basic option for starters. It isn’t pretty without some css work done to it at least, but it still saves you explaining aspects of new software to someone who really should be contributing their knowledge and expertise about the subject in the ontology, not become full time editors.

There is also a neater Apache Tomcat servlet, ontology-browser, which seems to work very nicely. The slight caveat there is that I don’t think we have a spare server lying around for running it on. Remote hosting could be an option, but is not ideal, given some of the nature of information that may end up in the ontology.

Finally, i stumbled across a presentation on SlideShare, talking about implementation of semantic import into Drupal. This has gotten me rather excited as I am a long time Drupal user, and like the extensibility of the system. There’s already ideas buzzing around on the potential power of combining importable ontologies directly with web-based presentation material of different instances. But I still have it to try, and it raises some similar issues in terms of server to the ontology browser – but still thought I would share the presentation here

After trying out a few ontology tools, I am slowly settling on Protege which seems to have a decent mixture of straight forward entity building with additional support for tracking object properties, data properties and individuals associated with classed . Given that this exercise is primarily about building a catalogue of activities whereby data from several underlying systems can be identified based on the relative ease with which it can be expanded is important.

Most of the ontology tools I have looked at are java based. This of course have the advantage of making them more distributable across operating systems, but can also have some quite severe memory implications – but so far Protege seems to work well without any signs of slowdowns.

In addition there are a number of plug-ins, which I am just starting to look into, but the ability to create forms based on the data definitions using the frames extension seems promising.

So at least for now, Protege seems like the choice for building a relatively simple ontology while keeping it scalable with the possibility of associating data directly.

As part of building a framework model for managing a rather varied collection of data, I’m seeking to develop a controlled vocabulary of activities to help facilitate better searching. However, I’ve now spent the best part of today rooting around in a long string of very impressive ontology managers that all seem quite capable of going the full mile if you want to develop semantic web applications. But I’m still coming up short on something a bit simpler – construction of a simple controlled vocabulary based on a relatively straight forward taxonomy. Maybe its me – still getting my head around the finer workings of semantics – pun intended