Functionally, this XML metadata standard closely resembles RDF. It models uniquely named things using triples, and has an "XTM" interchange format that can be read into "engines" (open source or not), then modified or saved as a triple store under an API (usually Java). Here are key differences from RDF:
(+) TMs are older by several years, and hence have a more robust toolset and bigger user community - mostly in Northern Europe. When I first got into them (2002), they were a more practical, deployable technology than RDF is even now, able to support queries, constraints, and inferencing in engine-specific ways.
(+) They have a robust "built-in" ontology, in which Topics, Properties and Names can be easily typed. They also do a much better job at handling subject identity in standard (versus app-specific) ways, and on data conversion and integration logic between themselves and other formats, including RDBs.
(+) BIG WIN: they replace RDF's binary Object relations with N-ary hashtables called associations, whose keys (called Roles) work like RDF Object Properties. For any Topic (subject), you can find all known associations in which it plays a Role, and all Topics that play a Role in each one of them.
(+) Every assertion in a Topic Map can also be "scoped" to a specific topic-set. Scoping has no counterpart in RDF, but it can be used in advanced apps to model "viewpoints", "sources", "theories", "sub-graphs", etc.
(-) It has no standard constraint language, and cannot use all of OWL - mostly just typing, plus domain, range, cardinality constraints.
Overall, XTM seems a more polished, mature, powerful, and rigorous declarative metadata language than RDF is. IMHO it wins handily on every technical comparison point, often by quite a lot, except the last.
HOWEVER, when one tosses OWL into the mix, RDF starts looking better. OWL adds a lot of new power to dress it up. One wonders if some way cannot be found to combine OWL and Topic Maps, and thus somehow get the best of both worlds.
ONE MAJOR USE CASE FOR XTM AND BIOPAX - ANNOTATED PATHWAY EXAMPLES
Topic Maps emerged from ISO not W3C, and they came out earlier, as the web-ized versions of back-of-the-book indices. They aim at human understandability, so their main power to help BioPAX likely lies in this distinction.
Associations, scoping, and better handling of identity and conversions let Topic Maps function as a higher level "authoring" and "editing" language for elegantly modeling things not easy to handle in RDF, but useful for pathways:
any type of collection, ordered or unordered;
any type of whole-part association, showing all named parts;
any pathway made of steps made of reacting participant-sets
scoped (conditional) variations on any of the models above
any annotation on any of the above, scoped by its source
Short, named subgraphs can be composed which directly model such things using a declarative Topic Map format and BioPAX terminology. This lets test cases be built quickly, reviewed and analyzed, and annotated to explore proposed changes. Testers can work in WORDS scripts, and/or directly in RDF/N3, and/or in Protege (dumping models into OWLDoc, which then gets converted):
![[WWW]](/wiki/modern/img/moin-www.png){kind=small}
http://www.lexikos.com/exhibit/cms/biopax/owldocycle.htmAn export module can then use OWL, plus the merged XTM and BioPAX concepts, to "compile" the whole assembly into semantically equivalent RDF export files for validation and other forms of reasoning.
Lexikos is now seeking funds for this sponsored R&D project for Q3 '05, to publish the required export machinery. Please email if you can assist:
The output process from this code will parallel, to a non-trivial degree, the same sort of compilation that takes Java source down into equivalent byte codes, or many other languages from some human-friendly format down to one easier to process by machine - in this case RSF. But here we deal only in declaratives, and prepare them for processing by an open set of semantic web reasoners.