Add further context to dietitian notes schema

[caufieldjh]

This is primarily about leveraging acronyms and adding more diagnostic criteria.

[caufieldjh]

Acronyms can be added with the gilda annotator:

• https://github.com/INCATools/ontology-access-kit/blob/main/src/oaklib/implementations/gilda.py
• https://github.com/gyorilab/gilda/blob/master/notebooks/custom_grounders.ipynb
  So the schema should include gilda:<path> among annotators, where is to a gzipped tsv of terms.
  But what does that file look like? See https://github.com/gyorilab/gilda/blob/master/gilda/grounder.py

That strategy isn't very flexible, though. How can we just nudge the LLM in the right direction?
Acronyms may need to be more contextualized in terms of the classes they're most relevant to.

[caufieldjh]

This may be less of a job for annotation (e.g., with gilda) and more of a job for the LLM itself, since acronyms have meaning extending beyond a simple key-value pair. One option is to set up a custom GPT w/ OpenAI but I'd prefer this to be more model-agnostic. This may suggest that OntoGPT could use some functionality to pass system prompts or some kind of initial instructions.

For now, these details can go in the schema.

[caufieldjh]

Improvements across the board here - but one recurring issue, more noticeable here due to involving more numerical value extractions than most: instances where the LLM adds extra formatting/punctuation tend to break things. Example:

  patient_weight:
    value: <45 kg; 35 kg>
    unit: AUTO:%3Ckg%3B%20kg%3E
    percentile: <46%; 6%>
    zscore: <-0.10; -1.55>

This will probably need to be handled in the SPIRES implementation itself as a post-processing step.

[caufieldjh]

Further updates in a future PR, but for now this will get results like (named entities omitted):

extracted_object:
  id: 45f2657e-36ee-4762-98f9-01665dd13e17
  label: Baby Girl A
  is_pediatric: 'True'
  is_preterm: 'False'
  patient_height:
    value: '44'
    unit: UO:0000015
    percentile: 8%ile
    zscore: '-1.43'
  patient_weight:
    value: '2.79'
    unit: UO:0000009
    percentile: 46%
    zscore: '-0.10'
  head_circumference:
    value: '35'
    unit: UO:0000015
    percentile: 92%
    zscore: '1.42'
  malnutrition_status:
    malnutrition_presence: N/A
    malnutrition_risk: N/A
    severity: N/A
    acute_or_chronic: N/A
    diagnosis: AUTO:N/A
    etiology: AUTO:N/A
  diet_supplementation:
    - AUTO:Starter%20total%20parenteral%20nutrition%20at%2080%20milliliters%20per%20kilogram%20per%20day
    - AUTO:glucose%20infusion%20rate%3A%205.5%20milligrams%20per%20kilogram%20per%20minute
    - AUTO:amino%20acid%202.8%20grams%20per%20kilogram%20per%20day
    - AUTO:dextrose%2010%25
    - AUTO:Trophamine%203.5%25
  nutrition_support:
    - AUTO:Custom%20total%20parenteral%20nutrition%20via%20peripherally%20inserted%20central%20catheter%20at%20dextrose%2012%25%2C%20amino%20acid%203.5%20grams%20per%20kilogram%20per%20day%20and%20soy%20medium-chain%20triglycerides%20olive%20fish%20oil%20emulsion%202%20grams%20per%20kilogram%20per%20day