Graham Moore: "Context is King!"
10.03.2008
From April 2 - 4, 2008 the Topic Maps 2008 conference will take place in Oslo, Norway taking a look at current developments and future prospects. Tassilo Pellegrini (SWC) talked to Graham Moore, Topic Maps expert at Networked Planet, about the semantic turn in IT, the power of context and the complementary nature between Topic Maps and the Semantic Web.
"Creating Context" is probably the crucial motto when it comes to realize the Social Web. How do you perceive this "Semantic Turn" in IT?
Context is King. If you are selling ads, they are worth more ‘in context’, if you are looking for new music you want it to be relevant to you. Context has become a key concern as consumers become more discerning and understand their needs. They now have an expectation that those needs should be met by the various content and information suppliers. To meet this demand the content owners are looking for ways, IT ways, to contextualize their content. This desire for content in context has two main consequences. Firstly, the need for a conceptual space into which things are placed. By its nature context cannot exist in isolation. This definition of context is about modeling a certain domain and critically, understanding where the user, the consumer, fits into this model. So far we aren't talking IT, we are talking philosophy and modeling. But, when it comes to building these systems it's important that we have technologies that are flexible enough to express these different domains and capture these contextual spaces.
"This is not just about contextualizing ads. This is about an effort for smarter information, smarter content."
Secondly, this semantic turn isn’t just about contextualizing ads, it seems to be part of a much broader effort for the enterprise to want deal in smarter information, smarter content. Simply put, you can do things of more value with contextualized information. For systems integration this is about aggregating multiple resources together into logical models from physical storage. This aligns information into a single holistic view that then supports better real time and high value business queries. Once again the user is a central part of that model. Similarly, for information management systems it’s about providing seamless logical models of concepts and content which allow users, in context, to find and be presented with the most relevant documents, emails, projects for their task at hand.
So we have user driven content consumption and enterprise driven data integration and improving information management systems, and while existing database technology can be used on a case by case basis to address these needs, IT people aren't like that and look to genericise, look to abstract and hence the 'Semantic Turn' in IT and the rise of semantic technologies.
The Topic Maps ISO standard has now been around for a couple of years and has developed - with ups and downs - a strong reputation as a viable standard for knowledge representation. What is the future of Topic Maps? Is it ready for the mass market?
Yes, I belive it is ready. In my opinion, readiness for mass market is about the completeness of the actual standards stack and secondly about its track record and adoption.
So let’s start with the standards stack. Many people still perceive topic maps as the XTM 1.0 standard from 2001 which contained a syntax definition, no formal model and little in the way of robust semantics. It had no schema or ontology language and no query language. Well, all that’s changed. The standards stack now looks like this; at the bottom we have the TMRM (topic maps reference model); this is a mathematical model developed out of a demand from academia to be more rigorous and a realization within the community that we needed something more low level upon which query and constraint languages could be built. On top of the TMRM is the TMDM (topic map data model); this is a higher level model that sits on top of the TMRM and is more inline with the original Topic, Association, Occurrence model that people naturally gravitate towards. The TMDM defines has its own data model but this is in turn defined in terms of the TMRM.
Next come the features standards as I like to think of them, TMQL (topic maps query language) and TMCL (topic maps constraint language). TMQL is a full featured query language. While its syntax constructs work at the level of the TMDM, the formalism of the language is defined in terms of the low level TMRM. Alongside TMQL is TMCL. This is a constraint language that defined constraints in terms of TMQL expressions. TMCL defines a mechanism to define constraints, and evaluation model for constraints and a set constraint types that do lots of the useful schema like constraints you need; such as indicating that a type is abstract, or requiring that all topics of type person have a least one name. As TMCL models all constraints as topics and TMQL expressions there is no new syntax for TMCL. However, to be more human friendly it defines a whole bunch of templates using CTM.
"I see a trend where large corporate organizations are looking to do this 'concepts and content thing' and Topic Maps is the choice."
So finally, we have CTM (compact topic map notation), which is a text based (non-xml) syntax for defining topics maps. It’s really designed to be easy to use for people and is heavily used in TMCL. So from a standards stack point of view I think Topic Maps is in a really good place for mass market. However, there are other key factors to consider; tooling and adoption so far.
Being a product vendor, I’m bound to say that the tooling has matured and is ready. The most notably way in which it has matured is that the product vendors have gone beyond the basic implementation of a topic map engine. They are now providing the value adding components on top which can exploit the features and value contained within a topic map. I also think that the technologies have had to grow up in terms of scaling and robustness.
And that leads me on to the last point about adoption. Topic Maps has always had a steady stream of adopters. However, a large number of those adopters were governmental, academic institutions or research activities. I see a trend now where it’s large corporate organizations that are looking to do this ‘concepts and content’ thing and topic maps is the choice. The world has matured since 2003, back then trying to sell topic maps was almost impossible, people still didn’t realize that links could live outside a document! Now, they realize they have a problem and that Topic Maps is the answer.
Topic Maps have traditionally been designed for inhouse use for knowledge management. The current trends go towards the web bringing up the issue of interoperability or data portability. How do Topic Maps deal with this issue?
As a standard, Topic Maps has some nice built in features that aid interoperability and data portability. A topic, a proxy for some thing, has a strong sense of identity. It uses URIs to indicate either that it is the proxy for some other resource (a file, a web page, a document in a CMS) or that it is the proxy for some subject. A topic’s subject identifier(s) references a page (that may or may not exist) that as unambiguously as possible describes the subject that the topic is a proxy for. What this means in practice is that if two parties agree (for me semantics is the agreement between two or more parties about what something is or means) that they want to ‘talk about’ the same thing, then they can both use the same Subject Identifier in their map. So, we have map A which has a bunch of topics in some way connected to this topic with the agreed identifier and map B which has another bunch of topics connected to a topic with the agreed identifier. I can now move those maps between different applications and have any processing keyed off the topic with the agreed identifier.
"Topic Maps has an inherent mechanism to merge two topics deemed to be the same. XML doesn’t have this notion of merging as part of its nature."
At this point people could say, ‘well, yes, but how is that different from two parties adopting the same XML namespace element for their vanilla XML?’ This is a good question. The difference is in what happens next; Topic Maps is always striving towards the subject uniqueness objective, one topic per subject. Thus it inherently has a mechanism of merging, whereby two topics, when deemed to be the same, are merged. Merging consists of taking the union of all properties of both topics and creating a new topic. You can’t do this generically with XML. It doesn’t have the notion of merging as part of its nature. This ability to merge topic maps together, and have topics that are the same (via the same subject identifier) merge, means that topic maps provides a great platform for EII (enterprise information integration), loose coupling of information management systems and cross publishing of content based on agreed concepts.
Recent developments in the field of Semantic Web seem to contest its stand. How do you perceive the relationship between these two technologies: competition or complementarily?
The RDF / OWL thing is interesting. To some degree there is a battle for mind share but if I had to broadly state, and I say broadly as the technical and expressive value of both is really on a par, where I see the two would be the following. RDF has more simple constructs and doesn’t really promote any particular notions. Topic Maps pulls up three of four different things and says, hey these are some of the most useful things for human - machine – machine - human interaction; namely, Names, Identifiers, Occurrences and Associations. RDF seems to have more traction in the business to business, machine to machine space, whereas topic maps seem to have been more successful in the human – machine – machine – human space and information management applications.
Looking forward I see a convergence of kinds, the low level TMRM and the RDF FOPL model can be seen as the basis for all that is stated in the higher levels. We have already seen through the W3C / ISO activity on RDF Topic Maps interoperability how they fit together. I can see this going further such that the tools have a very abstract internal representation that can deal with both flavors of semantic language.
What can we expect from the upcoming Topic Maps 2008 conference? What will be the big issues?
The great thing about the Topic Maps 2008 conference is that it’s all about adoption! Nearly all the presentations are about how Topic Maps has been used by a host of organizations to manage information, provider better access to information, to do EII. The community grows through seeing how others have adopted the technology and this is a really important event showcasing all that is great about the standard and the tools around it.
About Graham Moore & Networked Planet
Graham has worked for 8 years in the areas of information, content and knowledge management as a developer, researcher and consultant. He has held leading roles as CTO of STEP, Vice President Research & Development empolis GmbH and Chief Scientist Ontopia AS. He has been responsible for the development of knowledge management products including K42 Topic Map Engine, X2X Link Management Engine and e:kms knowledge suite. Graham is co-editor of the XTM 1.0 XML Topic Maps standard and IS013250-1 and -2 (Topic Map Data Model and Syntax), he is also co-editor of TMCL (Topic Map Constraint Language).
In his own words. "Along with Kal Ahmed I founded Networked Planet in 2004. We had been working with Topic Maps and content management for 4 years and were disappointed with the lack of traction such a powerful technology had gained. Networked Planet gave us a vehicle to develop topic map technology that more clearly delivered business value. Our main product is a topic map engine, TMCore, but we also have modules for Microsoft SharePoint and EPiServer CMS. I’m also co-editor of ISO13250 TMDM, and TMCL."
References: