Back again!

Yes my blog has been quiet for a while. I am battling cancer while trying to perform this research, unfortunately the cancer demanded more attention - I was hospitalized with pneumonia and a couple of other complications.

Today I want to explore the difference between the current standards and the context approach we developed in the 90's.

The context approach was based upon a simple model developed by Mike Mohammed. The model calls for the analysis of a domain, dividing its terms into a series of categories. These categories are:

• ENTITIES - otherwise affectionately known as "things" - if it is a noun it is an entity.
• ACTIVITIES - the "verbs" - actions that can be performed to change state.
• ATTRIBUTES - both entities and activities may have attributes that describe fixed and variable.
• CONTROLS - define rules by which activities may be initiated.
• RELATIONSHIPS - the most interesting part of the model describing many kinds of relationship. For example a person driving a car has a relationship to the car that is transitory.
  

These concepts are related in a simple model that enables the domain to be mapped to it's fundamentals. Models from multiple domains can be compared and mappings between domains produced much like a language translation.

By contrast the current standards approach requires no such structure. It would seem to me that most human language (may be all - I have not made an extensive study) divides the vocabulary into parts of speech: Nouns, Verbs, Adjectives and Adverbs. If this is basic to how we process language, it could also be a sound foundation for machine semantics.

My current efforts include:

1. Attempting to refine the model to provide a temporal component.
2. Attempting to define the model in current standards as basis for further development.

Both these efforts are more difficult than at first appears, however I feel the second is most likely to be the greater challenge.

Please contact me at roger.coates@webdirection.ca if you are interested in this research.

Semantic Research - RDF, RDFS, OWL

I still have much to learn in the world of semantic standards. However I have begun the construction of an ontology for genealogy. The ontology I am building will relate basic family ancestry to geographical information and historical events.

Anyone interested in this work please feel free to contact me on the WebDirection site - see the link on the left.

I am starting with a conventional ER model of the data. This will provide me a structured view of the data and enable some basic prototyping to ensure the approach is robust and consistent. Next I will build a context model for the information, and finally a standards based ontology.

In my research I have come across a number of semantic wiki tools these look interesting and I may include one of them on the webdirection.ca site.

As yet I am skeptical of the standards based approach.

I had an e-mail conversation with Mike Mohammed earlier this week - Mike is of the opinion that managing the relationships is the most challenging part of modeling and that a tool to do this would be a significant step forwards. I will be giving this some serious thought over the next couple of weeks.

Fundamentals - Its all about meaning

A quick survey of the emerging technologies for the semantic web shows and alphabet soup of new layers on top of XML. Unfamiliar acronyms and obscure terms like conceptual graph and ontology add to a bewildering new world.

Getting back to basics - it is all about providing meaning to information so that it can be interpreted by systems. In school we learned the basic parts of grammar for whatever our native language, we learned about nouns, verbs, adjectives and adverbs. It is a good idea to cast your mind back to these basics as we search for ways to create machine based meaning.

The human mind has developed an ability to interpret the same words in different contexts. For example bus can be a part of a computer or something you ride to work. If I say "the bus comes at 10:15" our minds interpret that I am talking about the type with wheels, whereas if I say "the bus runs a one megabit per second" the computer component is inferred.

In the mid nineties a small group of researchers working for an oil company launched a major initiative to create what we called the Context Information Base. There are many disciplines involved in oil and gas exploration, each of these has its own specific types of data as well as some common overlapping terms.

We based our development upon a model developed by Mike Mohammed, which we called Mike's Meta Model. The meta model facilitated mapping between contexts much as a language translator would between languages. We had some significant success with the approach. In fact the work was only stopped because we discovered that integrating the information  created significant changes in business process which were difficult to implement on anything other than a small team scale.

So for the next month while I grapple with RDF, RDFS, OWL and SWRL - I will be contrasting these approaches with our earlier research to see if there is any synergy.