formalize: use an LLM to match informal terms to terms in an ontology

In vjcitn/ontoProc: processing of ontologies of anatomy, cell lines, and so on

use an LLM to match informal terms to terms in an ontology

Description

use an LLM to match informal terms to terms in an ontology

Usage

formalize(
  informal_terms,
  ontology_terms,
  ontology_tags,
  ellmer_chatfun = ellmer::chat_openai,
  llm_model = "gpt-4.1-2025-04-14"
)

Arguments

informal_terms	character() vector of terms not necessarily found in ontology
ontology_terms	character() vector of ontology terms
ontology_tags	character() vector of tags for ontology terms, must be of same length as ontology_terms
ellmer_chatfun	function available in ellmer to connect to chatbot
llm_model	character(1) used with chat_openai in ellmer, defaults to "gpt-4.1-2025-04-14", or other models for other providers available through ellmer.

Value

A data.frame with columns informal_term, formal_term, similarity_score, and tag. Invisible attributes chat_tokens, chat_cost, and chat_provider are also present.

Note

Expects to have OPENAI_API_KEY set if an openai chatfun is used, or GOOGLE_API_KEY if, e.g., a gemini chatfun is used.

Examples

if (interactive()) {
 ctypes = c("tPlasma cells", "tMoMacDC", "tT cells",   # from Zilionis
  "tB cells", "tNK cells", "tNeutrophils", "Fibroblasts", "Type II cells", 
  "tpDC", "Endothelial cells", "tMast cells", "Smooth muscle cells", 
  "ND", "Club cells", "bNeutrophils", "bT cells", "bMonocytes", 
  "bNK cells", "bRBC", "bpDC", "bB cells", "bPlasma cells", "bPlatelets", 
  "tRBC", "Type I cells", "Ciliated cells", "bBasophils")
 cc = owl2cache(url="http://purl.obolibrary.org/obo/cl.owl")
 cloi = setup_entities2(cc)
 oname = cloi$name
 actual = grep("CL_", names(oname))
 oterms = as.character(oname[actual])
 otags = names(oname[actual])
 octy = formalize(ctypes, oterms, otags)
 head(octy)
 attr(octy, "chat_tokens")
 onto_plot2(cloi, unique(na.omit(octy$tag)))
}

vjcitn/ontoProc documentation built on July 5, 2025, 6:31 a.m.