Skip to contents

Latin-aware Botanical Name Matching

Source: vignettes/latin-aware-matching.Rmd
latin-aware-matching.Rmd

What “Latin-aware” means here

Botanical species names can differ because an adjectival epithet changes its ending to agree with the gender of a genus. For example, the Code gives Bromus mollis and Geranium molle as adjectival forms in different genera. That pattern can help a matcher find a candidate. It does not prove that the two names refer to the same taxon.

prepR4pcm’s Latin-aware fuzzy matching is deliberately narrow. It reuses pr_suggest_gender() to identify pairs in two common ending groups: -us/-a/-um and -is/-e. Those pairs receive an ending-aware spelling score, but the matcher allows them through only when the supplied taxonomic evidence has the same accepted identifier, provider, snapshot, and rank. If that evidence is absent, ambiguous, or conflicting, the pair remains available for review and is not returned as an automatic fuzzy match.

The helper can show the possible endings in a reference list:

pr_suggest_gender("Plant albus", c("Plant alba", "Plant album"))
#> # A tibble: 2 × 5
#>   input       suggestion  ending_group ambiguous requires_review
#>   <chr>       <chr>       <chr>        <lgl>     <lgl>          
#> 1 Plant albus Plant alba  us/a/um      TRUE      TRUE           
#> 2 Plant albus Plant album us/a/um      TRUE      TRUE

These constructed names demonstrate string handling; they are not claims about published taxa. pr_suggest_gender() keeps every matching alternative and marks every row as requiring review. It does not alter a reconciliation.

Fuzzy candidate generation starts with names whose genus spellings are close. This follows Erick Lundgren’s suggestion to work with genus evidence before species epithets. A strict genus cutoff would miss names transferred between genera, so a broader fallback is available when the same comparable authority identifies both spellings with the same accepted ID. The name score still reports spelling similarity; authority identity is kept separate from that score. A tree’s branch placement is not used as independent identity evidence.

The grammatical limit

The International Code of Nomenclature says adjectival specific epithets agree with the generic name’s gender. Nouns in apposition, genitive nouns, nouns with accompanying adjectives in the genitive, and adverbs keep their own endings. The Code also recognizes epithets whose grammatical category is ambiguous; resolving that ambiguity can depend on a published nomenclatural choice. These distinctions mean that a string ending alone cannot tell software whether an epithet is an adjective or whether it should change. This package does not parse Latin, infer adjective classes, or determine genus gender.

The ending groups implemented here are candidate patterns, not a complete account of botanical Latin. Greek-derived forms, other Latin declensions, irregular forms, historical orthography, and ambiguous published usages are outside this helper’s coverage. The rules above are grounded in the current Madrid Code online, especially Articles 23.1, 23.5, and 23.6.

What to do with a review candidate

Inspect the candidate and its authority evidence with reconcile_candidates() or reconcile_review(type = "identity_candidates"). If the authority does not establish a matching identity, check an appropriate taxonomic source yourself. Record a deliberate decision with reconcile_override() or reconcile_override_batch(); do not treat a grammatical-looking ending as a taxonomic crosswalk.

For the related spelling-only review helper, see the “Reviewing grammatical endings” section in the getting-started article.