Hello, I was hoping one of you good people might point me in the right direction. Please excuse my misuse of terminology, I am new to this particular facet of software design / modeling / etc. I am trying to get some high level feedback in order to help guide my learning process.
I work in the medical field and am interested in creating / training a model in order to explore automatic translation of the medical integration formats. To clarify a bit, medical data is transmitted amongst entities using several different transport formats (HL7, FHIR, CCDA, EDI X12, etc). These formats are highly structured. CCDA uses XML to define a reusable and shareable structure with which to share data, for example. Within these highly structured documents there is data that describes the medical event occurring - this data is often presented in the form of structured and curated medical taxonomies. For example, within the CCDA document there might be some elements that define the reason for your doctor visit, and it might use the ICD10 taxonomy to do this. Let’s say that the reason for my visit was V97.33XD: Sucked into jet engine, subsequent encounter. Which is a real taxonomic element by the way.
So what particularly do I want to know? A few questions:
are you aware of any projects or teams that are researching medical taxonomy translation?
When creating a training model, if the structure of the data is different than say human language, I.e highly structured, how does that influence design of the training model? These questions are difficult to ask because I feel as a newcomer I am lacking the correct terms. To clarify a bit more: in some of these data formats the position of the data within the structure defines what it represents. I would imagine that since this isn’t quite the same as natural language, even though it’s comprised of natural language, I would need to somehow explain to the model what the values mean in certain locations. I know there are labels and attentions as a concept utilized that might be what I’m looking for.
any recommendations to training/learning?
I hope this post isn’t too nonsense. To summarize m: in trying to translate between “languages” which are comprised of Latin characters and obey the general rule of English, but the pattern of the structure influences the meaning in ways unlike grammar. If that makes sense.
Thanks for reading, I really do appreciate it.