Extract a Key Resources Table from a manuscript

Reads a manuscript (PDF, JATS/NISO XML, DOCX, or text), extracts candidate resources with the deterministic regex engine (default) or an LLM, and returns them as a validated krt_tbl. Every result is provenance-stamped with the engine used.

Usage

extract_krt(
  input,
  engine = c("regex", "llm"),
  format = NULL,
  profile = "generic",
  llm = NULL,
  resolve = FALSE,
  existing = c("detect", "ignore"),
  title = NULL
)

Arguments

input: A file path or text string.
engine: "regex" (deterministic, offline) or "llm" (requires a configured provider and API key).
format: Optional explicit input format.
profile: Profile to assign to the extracted table.
llm: A krt_llm() configuration (for the LLM engine).
resolve: Whether to resolve identifiers during validation.
existing: "detect" to also parse an embedded KRT, or "ignore".
title: Optional title for the extracted table.

Value

A list with krt (the extracted krt_tbl), candidates (the raw resource candidates), existing_krt (a parsed embedded table or NULL), and report (a validation report).

Examples

res <- extract_krt("Anti-TH (RRID:AB_390204); FIJI (RRID:SCR_002285).")
nrow(as.data.frame(res$krt))
#> [1] 2