Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add grounding refiner option for SPIRES #361

Open
caufieldjh opened this issue Apr 1, 2024 · 0 comments
Open

Add grounding refiner option for SPIRES #361

caufieldjh opened this issue Apr 1, 2024 · 0 comments

Comments

@caufieldjh
Copy link
Member

(related to discussion with @enockniyonkuru re: MAXO extraction on Apr 1 2024)

SPIRES extraction does grounding recursively, but still doesn't always catch instances where the term to match is within a longer string, e.g.
lorem ipsum dolor TERM TO MATCH sit amet
Sometimes this is due to closely related texts not otherwise defined as synonyms, like vitamin supplementation (MAXO:0001129) vs vitamin therapy.

An additional refiner pass could assist with grounding by doing one or more of the above:

  • Performing an additional round of recursive searching, particularly in cases where the extracted string is longer than expected
  • Doing class-agnostic chunking with the LLM - essentially asking it to try again, but make it more/less specific.
  • Ditto, but with more traditional NLP methods, even just further tokenization
  • For ungrounded terms, replace common prefixes/suffixes with those more common in the source annotator. This may be better as a RAG approach but could also work in-context (i.e., instructions passed directly to the LLM) or as a post-processing step capable of recognizing repeated patterns among phrases. Sounds a bit like a curategpt thing though.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant