|
> Medical language is a reduced subset of
ordinary language, but
> also parts of the vocabulary of phrases are more precisely defined > than when they are in general use, and some of them are simply > not in general use. Langage generation (should) use 3 components : - the vocabulary (basic component) - the grammar (how to express meaningfull meta-concepts using the vocabulary) - the signature (don't know if it is the proper scintific word in english - it is in french) : that's the way some kind of people elaborate specific sentences ; I think it is what Adrian call "some of them are simply not in general use" > However it is (in my case and place) English, and should not be > replaced by a sequence of compositional codes. If I speak of "langage generation", I start from the assumption that you can generate sentences from something, that's to say you can express medical description in a way (structured langage) it can get generated in english or french, or german, etc (natural langages). Automatic translation systems are build that way : start from english, express sentences meaning in a "neutral - non ambiguous langage" - then generate french sentences (due to the 3 components - vocabulary, grammar and signature) Odyssee's trees (made with the Lexique) are a way to express that "neutral - non ambiguous langage", and I hope I will soon be able to show you that you can elaborate an echocardiographic report in English and print it in French - and only store it as a structured tree. People must understand that we are far away from coding. It is structured DESCRIPTION. As we can also describe pieces of knowledge the same way, we hoppe we can soon compare both patient's documents trees and knowledge trees to get decision support tools. > A vocabluary of phrases supplied so as to make them easy to > integrate into the notes, and with a link to an agreed meaning or a > note of the range of meanings assigned to them, would be useful > and can also serve as an educational tool. That's pure coding. How many sentences of the kind would you need to talk with people in the street (1 million or more) ? Medical vocabulary is - alone - as wide as everyday speaking vocabulary (30 000 terms in french). > What does the panel think about ontologies for pharmaceutical > and other prescribable items? I can only speek about Odyssee : currently the Lexique only deal with Common International Denomination (thats to say drugs components) ; We will try to extend it with a commercial base, but it is not easy since drugs databases are not built to "make sense". Typically, the job is to express semantic links of the kind : Aspirine 500 du Rhone -is a> Aspirine du Rhone -is a-> aspirine drug database drug database Lexique The main problem here is : aren't drugs databases alway country specific ? The Libre approach would be far more effective if we can collaborate, since it is a huge job. Regards, Philippe AMELINE Odyssee project www.nautilus-info.com |
