User:WhizoPopop

= Lexeme theories = The lexeme theories are based on some simple observations of the form and structure of professional reports. The theories lay the basis for LexeNotes medical note writing software.

Lexeme
A lexeme is defined as a unit of communication, divorced from the language used to express it. Lexemes can therefore be expressed in any language, or style including a unique computer code.

Lexeme Query and Response
A lexeme can be divided into a lexeme query, which defines the issue being addressed, and its associated lexeme response, which declares the substance of the lexeme. For instance the query “Frequency of headache?” might have the response “about once a week”. Lexeme queries have certain attributes that determine whether they will be used in a document or not. Lexeme responses may issue information that forms the basis for this determination

Context
Context is the information that is available before starting a professional note. In medicine, it includes the type of patient being seen, the clinical area where the interaction takes place, and the role of the clinician. For instance, a radiologist reporting on an X-ray in the emergency room  has a very different context than an oncologist seeing a return patent who is receiving chemotherapy.

Coherence
The theory of Coherence states that there is only one correct location for every lexeme query in the medical record. The proof of this theory is that if one removes a lexeme from a completed note, the knowledgeable clinician will be able to place it back in its correct location. The implication of this theory is that a lexicon of lexemes that covers the entirety of medicine can be constructed as single, ordered file.

Predicance
The theory of Predicance is that no lexeme query need be addressed in the medical note unless it is called for by an element of the context or by a previously selected lexeme response. For instance, the context of an internist writing a note concerning a history and physical exam will include a review of systems, specifically whether the patient has headaches. If the response is affirmative, the predicant indicating the present of headache is set to true, and that predicant permits the the query about headache frequency. The response “about once a week” might set another predicant for migraine, etc.

Level
The theory of Level is that no lexeme query need be addressed unless it is at least as senior as the hierarchical numerical level issued by the preceding lexeme response. A low value shows that the response does not call for a more detailed examination of the topic at hand, whereas a high number may permit a very detailed lexeme query to be addressed next.

The Lexicon
The lexicon is a computer file of lexeme queries listable in coherence order. Each lexeme query has an associated level, a list of predicants that permit it to be included, and its associated lexeme responses. Each lexeme response has associated text fragments, a unique computer code, a level and a list of predicants that will be set to true if the response is selected by the user. It is estimated that the entirety of medicine can be adequately addressed by a lexicon of a few million lexemes. The lexicon can be written by users who are authorized to be authors, lightly supervised by editors. The lexicon can be updated several times a day. Users can report errors immediately to editors.

How it All Works
The user sign-on procedure will set a number of predicants, and is being presented with a lexeme query concerning the clinical area and patent type, used to complete the require predicant to meet the context. Thereafter the user works iteratively, being presented with a question. and a series of possible answers. When the user selects the best answer, the system issues a text fragment associated with that answer to the output file, updates the level, and adds any new predicates to the system's predicant list. Then the systems hunts in coherence order through the lexicon seeking the first lexeme query that has adequate level, and at least one predicant that matches one in the system's predicant list.

This system has been studied in a clinical trial, and has been compared to OpenAI's GPT-4.