Le 05/04/2018 à 12:16, Thomas Beale a écrit : > On 02/04/2018 18:38, Philippe Ameline wrote: >> >> Actually, I don't think that I use grammar in an unusual way. If I do >> it technically, lets assume for the sake of the discussion that I am >> really talking about a grammar, ie a set of rules that allows you to >> interpret an arrangement of concepts as a discourse. Typically, a >> dependency grammar is not just a tree representation, but a tree >> representation where you take as a rule that the sons of an element >> qualify this element. Since every natural language sentence can be >> represented as a dependency grammar tree and vice versa, it is >> possible to assert that a dependency grammar is a sufficient grammar. > > Right - but the normal sense of 'grammar' is something that controls / > validates sentences made up of words so at least they have acceptable > structural forms, even if they say semantically nonsensical things. > The fils guides 'grammar' is supplying both levels - correct form (by > implication, due to /ordering/ of tree elements as you pass through > them) and valid semantics (due to the /content /of the tree elements, > thus preventing 'colon of stenosis' but allowing the reverse).
Let's imagine that there is no fils guide. A "patient record" is a graph of trees (means trees which nodes can be interconnected by typed traits, either to connect the description tree that implements a given document description tree, or to follow a given issue over time, etc). If you assume that this trees are organized as a dependency grammar and their nodes are filled using an ontology, you don't need anything else to read it, feed smart agents, etc. It is a story told using a structured language (ie a grammar and an ontology). Of course, as you mentioned, it is possible that it contains "wrong entries" like "colon located at stenosis". In the wild, it can be achieved by providing practitioners with just an interface where they can freely express themselves by building trees (this is the usual interface for encounter notes since it is fully non deterministic). Now, for many good reasons, we could want to guide the way (some/most) trees are elaborated (to ease and speed up the process, to make certain that the information we want to process will be "well put", etc). In deterministic areas, we can use archetypes. In semi-deterministic areas, we can use fils guides (a flexible way to guess and propose the next "sons" depending on current path). In my mind, fils guides and archetype are of different kind: an archetype is a flexible information schema and nodes that were "build using this mold" keep a link to it ; on the contrary, a fil guide is nothing more than a UI helper that makes a one step deep proposal (since, when validating a proposed son, you now are on a different path (previous one + validated node) and the system will try to find a fil guide for this path). Since the process is fully dynamic and local to the user (depending on the set of fil guides he uses) the nodes don't have to remember what fil guide they originate from. To sum it up, you can have a journey walking in well known areas (archetypes) and finding your way in the wild (tree filling interface). When in the wild, you can sometimes be presented with a "step wide carpet" (Fil guide) that helps you walk more comfortably (this process being iterative, you can "follow the carpet as it unfolds", but can also head on in another direction). > well maybe 'structural terminology' is a bad term; what I am really > talking about is /models of possible content/ (possible utterances). I was mainly talking about the extra structural elements such as "Entry", etc. Besides, if "models of possible content" are very important inside the deterministic area, it would be pretty limited to have the only alternative "model of free text". If only because, if you provide users with a structured language, you will also be able to detect that they enter an area where you can present them with a model. When I wrote that a fil guide only makes a "one step ahead" proposal, I forgot to mention that it can also trigger an archetype (hey, since you are mentioning blood pressure, why not simply fill this form?). > > yes, there is no doubt that the way you engineered /fils guides/ > achieves this very well, and we have things to learn in terms of > bridging the gap between linguistic expression and structural > expression - for now, openEHR has no 'system' to do the former, it is > just done /ad hoc/ by those who want it. Fils guides are fit in a semi-deterministic environment and only when there is a reference ontology available (because it compares user's current path with semantically similar expert designed paths). I really hope we can cooperate in this direction. Philippe
_______________________________________________ openEHR-technical mailing list openEHR-technical@lists.openehr.org http://lists.openehr.org/mailman/listinfo/openehr-technical_lists.openehr.org
