Classify disagree rows as auto-resolved, suggested, or leave as disagree

Runs after compute_similarity_scores(). For each disagree row, collects per-candidate scores and applies threshold logic (D-01/D-02/D-03):

• Auto-resolve: best score >= 0.95 AND gap >= 0.15

• Suggest: best score >= 0.70 (but not auto-resolve eligible)

• Leave as disagree: best score < 0.70

Usage

classify_auto_resolve(
  resolution_state,
  enrichment_cache,
  dtxsid_cols,
  column_tags,
  auto_threshold = 0.95,
  gap_threshold = 0.15,
  suggest_threshold = 0.7
)

Arguments

resolution_state

Data frame with consensus_status, similarity_score, dtxsid_, preferredName_, rank_* columns

enrichment_cache

Tibble with dtxsid and (optionally) synonyms columns

dtxsid_cols

Character vector of dtxsid column names

column_tags

Named list (col_name -> "Name"|"CASRN"|"Other")

auto_threshold

Numeric, minimum score for auto-resolve (default 0.95, per D-01)

gap_threshold

Numeric, minimum gap between best and second-best (default 0.15, per D-03)

suggest_threshold

Numeric, minimum score for suggestion (default 0.70, per D-02)

Value

resolution_state with updated consensus_status, .pinned, .resolution_method, .resolution_reason, and .suggested_column columns