Two issues are being conflated herre. 1. Standards The sole purpose of a standard is to guarantee interoperability but, to achieve this, the standard itself must satisfy certain criteria. These criteria are well illustrated by documents such as the old Whitworth standard for nuts and bolts. This defines the way in which the dimensions of nuts (diameter, threads per inch, etc.) and those of bolts must be specified, this definition being both: a) formal enough for the manufacturer to be able to prove that his products satisfy their specifications and b) necessary and sufficient to guarantee to the purchaser of a nut or bolt that it will fit the bolt or nut she already has. Unfortunately, few, if any, of the 'standards' in Healthcare meet these criteria. 2. Language Languages, both natural and formal, can be described at a number of levels depending on the purpose of the description. For mere transmission of the symbols, we use lexical definitions. To identify significant strings of symbols (i.e. to parse), we use syntactic definition. To interpret the intended meaning of such strings (i.e. to compile), we use semantic definition. These levels suffice for the formal languages. For natural languages, we need a further level of description that includes the social context of the utterance, this being known as 'pragmatics', a term was by Charles Sanders Peirce who also named the entire field of language description and analysis, at all those levels, as 'semiotics'. Computational techniques have made huge inroads into semiotics in recent years, particularly up to the level of semantics, but pragmatics is still a dark art. Ontologies (another term coined by Peirce, who also did the original work on ontology definition, as Sowa himself acknowledges) provide some leverage but, since there can be no universal ontology, and the composition of disparate ontologies is not computationally feasible, only those communities who can identify with a common ontology can benefit, and as this identification is always merely temporary, later dissatisfaction with its implications is inevitable. I hope this is not seen as too pessimistic. There are effective ways of dealing with these problems but only if we learn how to include the subject, and the subject's models, in our models. The 'objective' forms of analysis that have been so successful in engineering simply will not suffice.
Quoting Gerard Freriks <gfrer at luna.nl>: > Hi Charlie, > > I agree. > This topic is not about HL7 and/or EN13606. > It is about the logical, semantic and technical aspects of semantic > interoperability. > > I like to think about problems using simple, time proven, solutions > and ways to deal with complexity. > One magic solution for everything is impossible. > We humans use the dictionary to describe the meaning of words. > Using a syntax we produce sentences. > With common knowledge in our heads we know what is relevant and makes > sense. > We express what we need to express in a context. > > The dictionary will not tell us what to document. > We need a way to capture what we want to express. > We all use documentation patterns to express things in a common way. > > Mental exercise: > - Documentation Pattern: " Once upon a time there was a Princess" > We humans know that it is the documentation pattern for a fairy tale. > Will the ontology be able to 'know' this? > Probably not. It will assume: there was a princes, there was a time, > there was a place. > > I see the need for six orthogonal levels (models). > 1- A structure describing knowledge = the Ontology > 2- Words to express knowledge = Coding system > 3- Something else > 4- A structure to assemble words into sentences > 5- A structure to assemble sentences in documents > 6- A structure to store meta-information for archiving purposes, > versioning, etc, etc. > > Without 3 we are able to produce correct sentences and collect them in > documents. > This does not guarantee that we produce relevant sentences in a > particular context. > It does not guarantee that sentences produced make sense; they can be > non-sensical. > Even when they are correct the documentation pattern causes the > interpretation to change completely. > Using 1 we (and IT-systems) will find out that it is nonsense and not > relevant in a context. > Therefor we need a structure so users can express want they want to > express. > This level three are Archetypes/Templates. > > Level 3 is the Documentation Pattern where context, processes, humans > interact with systems and use all the other layers to document, > archive, exchange and re-use heir data and information. > > At level 3 we must know how technically we can refer to codes from > coding systems. > I know that we have not a universal way to refer to codes and coding > systems. > > Do we have a worldwide agreement how we refer to a coding system? > Do we have a worldwide agreement how we refer to a specific code from > a coding system? > Do we have a worldwide agreement how we refer to a defined subset of > codes from a coding system? > How do we deal with the variable structure of each code? > Do we have a worldwide agreement how to process the presentation > labels and descriptions? > Do we have a worldwide agreement how to express inclusion and > exclusion criteria (in the case of classifications for example)? > Do we have a worldwide agreement how we deal with the language in > which code and coding system related items are expressed? > Do all standards, systems, specify all this in universal way? > > Gerard > > > -- <private> -- > Gerard Freriks, MD > Huigsloterdijk 378 > 2158 LR Buitenkaag > The Netherlands > > T: +31 252544896 > M: +31 620347088 > E: gfrer at luna.nl > > > Those who would give up essential Liberty, to purchase a little > temporary > Safety, deserve neither Liberty nor Safety. Benjamin Franklin 11 Nov > 1755 > > > > > > On Apr 23, 2009, at 9:39 AM, Charlie McCay wrote: > > > I would agree that there are limits to the utility of such mappings > > - indeed it is to explore such limits that we are engaged in this > > thread. > > This is a serious area, and openehr, 13606, and hl7 all have > > mistakes and successes, (and differences and similarities). We have > > differing perspectives on those, but let's try to put that to one > > side and address common themes in this thread. > > I agree that there is a difference between language and ontology. I > > am less convinced that to serve clinical system interoperability the > > distinction can be maintained absolutely. At one level there is the > > blurred boundary between terminology and structure, and at another > > there is the safe automated reuse of entries/clinical statements - > > something that happens and for which we need a better understanding, > > with entries being treated as semantically independent. I beleive > > that ontologists have much to contribute to this area. > > I share with Seref a desire to understand why the research work is > > not getting into practice. If it is not addressing the practical > > questions then I move on to ask what work is. > > > > My interest is in asserting the relationships between standards > > relevant to interoperability. I beleive that there is value in > > seeing what is stopping this happening, and whether the cost of > > addressing some or all of those hurdles would be justified > > > > All the best > > > > Charlie > > -- __ Prof Bernard Cohen, Dept of Comp Sc, City Univ, Northampton Sq. London EC1V 0HB tel: ++44-20-7040-8448 fax: ++44-20-7040-8587 b.cohen at city.ac.uk WWW: http://www.soi.city.ac.uk/~bernie "Patterns lively of the things rehearsed" ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.
