Review semantic suggestions in the console

Builds a re-runnable review queue from suggestions that already exist. One entry per unfilled semantic slot, plus NuSEDS crosswalk-prefilled code slots that still hold their original prefill, each with its ranked shortlist and the exact accept_suggestion() call that decides it – printing that call is the feature: paste it into a script and the decision becomes reproducible, which the spreadsheet workflow this replaces never was.

Usage

review_semantics(
  x,
  include_filled = FALSE,
  max_candidates = 5L,
  columns = NULL
)

Arguments

x

A written package path, a dictionary carrying the semantic_suggestions attribute, or the artifact list returned by infer_salmon_datapackage_artifacts().

include_filled

Logical; if TRUE, also queue slots that already hold a final (non-REVIEW:) IRI. NuSEDS crosswalk-prefilled code slots with a saved shortlist are shown by default until decided. Defaults to FALSE.

max_candidates

Maximum candidates shown per slot. Inf shows all.

columns

Optional character vector restricting the queue to these column names.

Value

An ms_semantic_review tibble subclass, one row per candidate, in the order the ranked producer emitted them. Carries the package path (when there is one) as the review_path attribute.

Details

This never contacts a network or an LLM. It reads the semantic_suggestions attribute (or semantic_suggestions.csv) that suggest_semantics() / create_sdp() already produced. When those suggestions carry LLM review – only possible if they were generated with llm_assess = TRUE – this surfaces it; it never generates it.

See also

accept_suggestion(), reject_suggestion(), apply_sdp_semantics()

Examples

dict <- tibble::tibble(
  dataset_id = "demo-1",
  table_id = "spawners",
  column_name = "spawner_count",
  term_iri = NA_character_
)
attr(dict, "semantic_suggestions") <- tibble::tibble(
  dataset_id = "demo-1",
  table_id = "spawners",
  column_name = "spawner_count",
  code_value = NA_character_,
  dictionary_role = "variable",
  target_scope = "column",
  target_sdp_file = "column_dictionary.csv",
  target_sdp_field = "term_iri",
  target_row_key = "demo-1/spawners/spawner_count",
  label = "Spawner Abundance",
  iri = "https://w3id.org/smn/SpawnerAbundance",
  source = "smn",
  ontology = "smn",
  definition = "Mature salmon returning to spawn.",
  score = 4.9
)

review <- review_semantics(dict)
review
#> ── spawners · spawner_count · variable ───────────────────────────────────────
#>    field:   column_dictionary.csv · term_iri
#>    current: <blank>
#> 
#>    [1] Spawner Abundance   smn   score 4.9
#>        Mature salmon returning to spawn.
#>        https://w3id.org/smn/SpawnerAbundance
#>        review <- accept_suggestion(review, "spawner_count", "variable", rank = 1)
#> 
#>        review <- reject_suggestion(review, "spawner_count", "variable")   # no candidate fits
#> 
#> ── next ──────────────────────────────────────────────────────────────────────
#>    0 of 1 slot decided.
#>    Paste the calls above into your script, then write the decisions:
#>    apply_sdp_semantics(<package path>, review)
#>