Comparing Topic Maps and RDF
The Topic Maps and RDF technology stacks are quite complex, and it's not easy to see how the various pieces compare. So I thought I'd write a little bit about this. It was Trond's comment that triggered this, but I've seen no end of people struggle with this. I've written about this before, but I think it's possible to do a lot better. So, here goes.
The big picture
The big picture |
The figure above illustrates the basic relationships between the various pieces of the puzzle. How this works is best expressed with some examples:
- Let's say you want to express that "Lars Marius is employed by Ontopia". For this you would use either Topic Maps (ie: TMDM) or RDF (ie: the basic triple model). That's because these are the basic data models in which all information is expressed, and the statement is what I'd call "instance information", or "application data".
- If you wanted to say that "employment must relate a person to an organization", on the other hand, you would have to use TMCL (on the ISO side) or RDFS (on the W3C side). (You're probably wondering about OWL now, but we'll return to that.) This is because we are now talking about constraints, and no longer about instance information.
- Now, if you wanted to find "all people employed by Ontopia" you would have to use TMQL (ISO) or SPARQL (W3C). These are the standard query languages in the two technology families.
- Finally, if you wanted to take the statement about Ontopia and me and export it from your application for transfer somewhere else (or just plain storage) you'd have to use XTM (with Topic Maps) or RDF/XML (with RDF). These are the standardized XML-based interchange formats on the two sides. (You could also use LTM or n3, but these are text-based, non-standard, and optimized for human readability.)
So where does OWL come into this? Well, in one sense OWL is just the big brother of RDFS, which lets you say lots of things that RDFS isn't powerful enough to let you do. However, OWL is not really a constraint language, but a language optimized for describing ontologies so that you can do inferencing on top of the instance data based on that information. I've blogged about this difference before, so go there if you want to know more.
The questions
With this out of the way it becomes easier to answer the questions Trond asked:
I am curious as to why the XTM community "always" (papers, mailing lists, etc.) compare XTM to RDF/XML and not to e.g. OWL Lite.
The reason should be quite plain now: these are the two standard XML-based interchange syntax in the two technology stacks, so RDF/XML really is the natural equivalent of XTM on the RDF side.
RDF/XML lacks the ability to define new classes and relations / ontologies, which is the purpose of RDFS & OWL (as you would know). Is it really fair to compare XTM to a technology which was not designed to do this, when alternatives / more expressive competing technologies do exist? I do understand Ontopia's (and other companies') motivation behind this, but is the XTM community really better off by ignoring OWL as a competing language/tool?
The simple answer to this is that TMCL is the natural technology to compare OWL with. Of course, the problem is that TMCL is not yet really here yet. However, I have reason to believe that the next draft will take us to the point where TMCL really is here, even if not 100% finalized yet.
If I was new to XTM and in need of a technology in which to define my ontologies, I'd most def. not compare XTM to RDF/XML, but to OWL (which, according to Anne Cregan, may also be used in order to express Topic Maps).
Well, you can't define your ontologies any more in XTM than you can in RDF/XML (or in Topic Maps than in RDF, which I think Trond really meant). To be able to define an ontology beyond just creating the various concepts and wiring them together you really need TMCL.
As for being able to express Topic Maps in RDF, what Anne Cregan really did was to create an RDF vocabulary for expressing Topic Maps. That is, she made a set of RDF properties and RDF types so that Topic Maps information could be represented in RDF as triples. I did the same thing way back in 2002, but I knew even then, that this was not the right way. The thing that Anne did, which I did not, was to describe her vocabulary with OWL, which let OWL work out some of the computed properties of TMDM for you.
I also recently got an email from Patrick Durusau where he wrote:
I am in a conference call presentation where topic maps were described as being "less expressive" that RDFS/OWL.
Patrick disagreed with this, but I'm afraid I do not, if you interpret "expressive" in the sense it's used in logic, where it refers to the amount of logical inferences that some language or representation allows you to draw on the basis of other data. RDFS/OWL really do allow more inferences to be drawn than the basic Topic Maps standard (TMDM) does. I think it's likely that they will also allow more inferences than TMCL, but that TMCL will reduce the gap significantly.
Similar posts
Ontologies: validation or reasoning
Information about the structure of an ontology can be used for two different purposes: either for validation or for reasoning, and this is a distinction that it seems most people are not aware of yet
Read | 2006-10-15 19:37
Semantic Web adoption and the users
A hot topic at ESWC 2013, and many other places besides, was the issue of Semantic Web adoption, which after a decade and a half is still less than it should be
Read | 2013-10-18 15:43
The web's identity crisis and httpRange-14
URIs are used to refer to both information resources (which are downloadable over the net) and abstract concepts and physical objects (which are not)
Read | 2007-10-08 08:54
Comments
rho - 2007-01-12 11:19:40
BTW, SPARQL bounced back to the WG and did not go for recommendation. I sort of suspect that the 'logician lobby' caused this. Now there is an "entailment regime" as part of the spec, which implies for me that SPARQL is now sitting _on top_ of that stack.
Maybe one should also mention that the RDF model is better comparable to something low-level like TMRM.
Trond - 2007-02-13 15:35:47
"Well, you can't define your ontologies any more in XTM than you can in RDF/XML (or in Topic Maps than in RDF, which I think Trond really meant). To be able to define an ontology beyond just creating the various concepts and wiring them together you really need TMCL."
True, but if we think of an ontology as simply making the concepts (i.e. objects and relations and/or constraints) of a domain explicitly available for machine processing, then Topic Maps does indeed provide more constructs then RDF, doesn't it? In TM (and XTM) you can, for example, express new topic and association *types*. In TM/XTM you may also organize such concepts in taxonomies. On the RDF-side, you have to turn to RDF Schema for similar constructs (rdfs:Class, rdfs:subClassOf, rdfs:subPropertyOf, etc.).
Therefore, I still don't quite agree with "always" comparing TM (XTM) with the "lowest level" of the W3C Semantic Web technology family, i.e. "plain" RDF (RDF/XML), instead of e.g. RDFS or OWL Lite (although I do get your point(s)).
Of course (as you note) you might want to define the ontology "beyond just creating the various concepts" and hence need the constraining constructs of OWL (owl:FunctionalProperty, owl:ObjectProperty, owl:unionOf, owl:oneOf, owl:minCardinality, etc.), or TMCL.
Anyways, I sure hope TMCL will see the light of day in not too long. The XTM syntax was released in 2001. We're now in 2007. I'm sure the use of TM/XTM would spread to new areas if a standard constraint language for Topic Maps did exist, 'cause right now it *is* far less expressive than OWL.
That being said, I do not believe that OWL is superior to TM/XTM no matter what. One might not always /want/ all the expressiveness of OWL (something the W3C acknowledged by dividing OWL into three sub-languages). Besides the logics and implementation issues/costs, it is, for instance, not a given that e.g. web applications are made more usable by introducing a "strict" information structure ... unless used by the Dept. of Homeland Security or something like that. After all, TM was made for navigating information (and companies probably want their customers to be able to find what's out there), OWL for supporting reasoning and a "true" semantic web and -agents - as seen by the W3C. They are both good at what they do, but adding TMCL to the TM-side would, of course, make it even better.
Lars Marius - 2007-02-15 16:11:00
Topic Maps do provide more constructs than RDF, in that Topic Maps have subclassing (which is in RDFS), scoping (which isn't in RDF at all), and a few other things.
However, you can't define topic/association types any more in Topic Maps than you can in RDF. You can express and use them without defining that they are topic/associations types, but that's the same in RDF. The only difference is really subclassing.
As for TMCL, the model has really stabilized now (we think), and the editors are working on a syntax for it. I think after the Oslo meeting in March we should actually have the first TMCL version you can actually use. There may be 1-2 more iterations to go after that, but we should really start seeing progress.
I definitely agree that the appearance of TMCL will affect the adoption of the technology.
As for what you write about OWL: I agree completely.
Trond - 2007-02-16 10:24:29
"However, you can't define topic/association types any more in Topic Maps than you can in RDF. You can express and use them without defining that they are topic/associations types, but that's the same in RDF. The only difference is really subclassing."
Of course. Sorry about that...being a bit hasty there, I guess (hope :D).
My point was about subsumption which, in my world, means that it's more "fair" to compare Topic Maps to RDFS (which also supports things not supported by TM, like rdfs:range and rdfs:domain) instead of RDF which doesn't even support declaring new classes or class-subclass relationships.
Good news about TMCL :)
Omar Diego - 2009-08-08 12:45:49
Great comparison between Topic Maps and RDF, it gives the big picture of the comparison. However something that is pending in the big picture is the Topic Maps Reference Model, the equivalent would be RDF specification?
vuelos - 2010-03-30 15:33:10
I agree with Trond. Comparing T Maps vs RDFS is like compare apples vs peaches.
Although this is great work, comparing TMaps to RDFS would have been more fair.
Enesi - 2013-09-05 13:18:20
Trond and all other contributors...a good job, well done. It might interest you to know that i'm currently working on my master thesis using XTM to develop Ontology for the (news)domain. However, I must confess, that i'm under pressure to search for reason(s) why I decided to use the technology (XTM) and not RDFS or OWL for my ontology...
Please, can any one help me out!!!
Enesi - 2013-09-07 07:44:41
Please, i'm still expecting respond/reaction to my mail (comment).
Warmest Regards...