This shows you the differences between two versions of the page.
sbvr-topics [2018/12/30 01:37] (current)
|Line 1:||Line 1:|
|+||====== SBVR Research Topics ======|
|+||* SBVR inspired a lot of European academic research that was mostly based on supporting and using its formal notational style for various applications. At the same time, the dozen or so ‘implementations’ utilized the basic ISO 1087-1 model and SBVR linguistic and logic terminology as a (pseudo-) formal semantics for their own languages. (I have my own, published in 2011.) There are also dozens of academic ‘controlled English’ languages that produce some kind of formal representation, and have no explicit relationship to SBVR. In short, this domain is no longer very interesting. (And the big guns in commercial AI are trying to interpret true natural language on an ever broader scale.) //Edward Barkmeyer//|
|+||* The formulation of a clear meta-model for unambiguous ‘controlled natural language’ concepts is apparently missing from the literature. SBVR’s use of its own language to define its concept system maps (using a section in the standard to guide the mapping) to a bizarre metamodel, which is only partly depicted in the specification (owing to the indirect approach and other factors). But SBVR makes no effort to model the grammar of its ‘Structured English’. So there is probably room for research in producing usable formal UML/MOF models of the vocabulary concept system and of the ‘Structured English’ grammar concepts. There is, of course, a possibly very close relationship to formal linguistic analysis models used for natural language interpretation, but those are burdened by the eccentricities of natural language. //Edward Barkmeyer//|
|+||* There is no formal mapping from the ‘structured English’ grammar to the (not clearly well-founded) formal logic language defined in the SBVR specification, or to any other formal logic language, such as CLIF or OWL. There is a table of concept/term equivalences (developed primarily by Elisa Kendall, I think) for the basic SBVR concept system to CLIF and OWL (both of which are necessarily partial). Other controlled English language publications have formal mappings to some formal semantic model, which may or may not have a well-defined formal logic interpretation. (It is not clear to me that any purported ‘SBVR implementation’ actually has and uses such a mapping, except for simple subsets. Some, for example, map concept systems to UML class diagrams.) I had hoped that the academic research spawned by SBVR would have gone in this direction. //Edward Barkmeyer//|