Judge lexical candidates with a provider

Usage

ducksemantics_judge(
  text,
  mentions,
  runner,
  conn = NULL,
  prefix = "semantic",
  graph_context = NULL,
  instructions = ducksemantics_default_judgment_instructions(),
  prompt_builder = ducksemantics_judgment_prompt,
  parser = ducksemantics_json_judgment_parser(),
  record = !is.null(conn),
  model = "semantic-runner",
  ...
)

Arguments

text: Source text.
mentions: Candidate relation.
runner: Prompt provider.
conn: Optional DBI connection to record judgments.
prefix: Table prefix.
graph_context: Optional graph context.
instructions: Explicit judgment instructions.
prompt_builder: Function building the prompt.
parser: Judgment parser provider.
record: Record result when conn is supplied.
model: Provider/model identity.
...: Arguments passed to prompt_builder.

Value

Source-grounded judgment relation.