Build the Semantic Web

I will install a semantic wiki on www.wikiworld.com and build tools for collective ontology building both passive and active.

Since my last entry, I’ve written two! more application profiles. I can’t believe I am doing this actively. This is awesome.

The other entries are doing it from a delivery point of view, but I do this more from the content interoperability point of view. It’s amazing to see the content that’s being created by the business folks in my organization (a really big distributed organization) being federated. Most of the folks don’t get taxonomies, IA and all the stuff that goes on to make this happen, but they DO get finding their stuff. And that’s ultimately what matters to me.

And I’m helping to make it happen for them. Wowza.

..before any of the stuff I’m now learning about will be in mainstream use. Many people say Semantic Web will never catch on. I wonder what are they thinking – what is the alternative? We stick to keyword-based search engines till the end of time?

No. Web will evolve. A lot. It’s inevitable. It’s true that semantic technologies have a lot of room for improvement, and that it will require a lot of manual work for this thing to work, but when it’s done.. just imagine the results! That keeps me going.

That’s what I’m into these days

So I just wrote my very own metadata schema. Yes, that’s right. I found a schema that wasn’t quite what I needed, took a little bit from it, and a little bit from another, and then some new information from my head and presto. New metadata schema.

ANd you know what? I have actually put it into production. Yes, that’s right! I have actually described documents and searched effectively using this schema! AND! It gets better… I have mapped it to Dublin Core and other schemas. God, so amazing to actually have done this.

... Now to namespace it. But wow! My very own Real Life schema!

... I’m such a friggin’ nerd. (See my other ToDo action item)

I’m co-developing a robust wiki that has enabled Semantic MediaWiki. It’s at Centiare.com, and I hope that it will become a huge database of personal and enterprise Directory listings. The cool thing we’re doing is that we’re assigning “ownership” of space to contributors.

That way, if you want to create an awesome, semantically tagged listing about your IT consulting firm, you can. And nobody else will be able to edit or vandalize it. But, if you’re writing on a non-ownable topic (e.g., the Rocky Mountains, rainfall, or dogs), it’s traditional, community-edited space.

I encourage folks to check it out, if you’re truly interested in “building the semantic web”.

I’ve create a FAQ about this:
http://protege.cim3.net/cgi-bin/wiki.pl?ProtegeScriptTabFAQruby

I’m considering getting together some kind of tutorial session on semantic web technologies for OpenPublish 2006 – a basic introduction to the concepts, and building a quick application atop a triple store in less than 40 minutes.

If anyone has some suggestions about how I should go about it or things to make special note of, please feel free to drop me a line or comment on this post!

As my other entry indicates, I want to run RDF inside Ruby on Rails. The ultimate reason for this is to “drive” the web application development process with OWL-DL. As a result, instead of Model Driven Architecture (MDA), there will be Ontology Driven Architecture – and this just feels right to me. After years of begging and pleading for the relational model to perform for me (it’s fundementatly flawed, I think), an architecture based on a precise model of the entities and relationships is going to change everything.

I’ve gotten a little bit further along in my goal. I had to switch temporarily over to linux (I’m using Cygwin). I needed to understand the fundementals of Redland and how it interacts with Ruby. I’m keeping a journal of this activity over at MySpace.

Lots of people have looked over Deep Integration of Ruby with Semantic Web Ontologies . Well, so have I. That article got me off the J2EE track and onto the RoR track. The more I started reading about the Ruby language and the Rails framework, the more sense it made. So I’ve decided to learn as much as I can about the technology.

One of the reasons for selecting Ruby is that the company I work for has a cultural bias against anything Java – and trying to inject J2EE in that environment may not be wise. I’m fairly confident that J2EE will probably rule the world and most companies won’t have a choice but to move to that platform but right now the ideology is firmly rooted in Old School Microsoft.

Ruby seems like a more pragmatic approach that may be more palatable and easier to sell to corporate management. It can be easily described as a scripting language like JavaScript (no problem there) and ground in academia. The hope is that there won’t be the typical knee-jerk reaction of proposing a java solution.

Actually, Rails and Ruby don’t have a thing to do with the Semantic Web. But the Obie’s article serves as a great starting point. The idea is simple: Rails is a framework based on the Model-View-Controller (MVC) paradigm implemented using Ruby as the language. The W3C has invented the Web Ontology Language . OWL is a metadata and descriptions language that will form the ontology layer of the semantic web. It should be possible to use the Ruby language to wrap around OWL ontologies and form the “Model” part of the MVC architecture of a typical web application.

One very interesting quality of the Ruby language is that unlike Java, Ruby is not strictly typed. Creating dynamic code should be much easier to accomplish. Ideally, a thought worker would describe the problem domain in OWL using a tool such as Protégé . The OWL could be then be uploaded to the Rails web application which would dynamically wrap around the OWL classes and properties and automatically provide context services (behavior) such as the acquisition of individuals. In other words, upload the OWL “schema” if you will, then web pages would magically appear allowing you to add information to a knowledge base.

If this is the goal, one thing we have to decide on next is what the underlying data store will look like. I’ve been investigating the Redland frame work and it looks promising. I was hoping to use Sesame triple store (which I already had up and running). Unfortunately, Sesame is a web application deployed to J2SE servers which might make the whole thing more complicated than I’m willing to deal with at the moment.

The interesting thing about Redland is that it understands SPARQL which will be very important for doing some very dynamic processing. SPARQL is currently being weaved into Protégé which already included SWRL rules support. Not sure of the direction of Redland for SWRL but something else may come along.

Ideally, the underlying data store would look something like this: Generic RDBMS (MySQL?) underlying the whole thing. XML Database perhaps tied into or some how integrated with a triple store. But how do we hook up between different remote stores? This I’ll have to think about some more. The idea of different web applications tied together via interlinked data stores sound positively incredible to me. But it’s probably a long way off.

So, were does all of this end up? It seems to me that if you have dynamically generated web pages based on a rigorous description language like OWL, throw in a little SWRL rules and some well thought out services, you’re on the right road to artificial intelligence.