Two entities (note: Entity) share a title or an alias (note: Alias) once spellings are compared: same form (52 passages, no note) after lower-casing, accent (23 passages, no note) stripping and singularisation. The recognition dictionary (note: Dictionary) keeps both, every occurrence (note: Occurrence) of the form links (note: Link) to each entity at half the confidence (note: Confidence) it would have with a single target (15 passages, no note), glossary (28 passages, no note) entities first, and the finding (note: Finding) names the form and the entities. When the two are different things, distinct titles or aliases (11 passages, no note) settle it; when they are one concept (3 passages, no note) seen from two sides, as the glossary term (note: Term) and the business object (28 passages, no note) called "Entity" in this wiki, both stay and a ## Not to be confused with section (note: Section) tells the reader (note: Reader) which is which. The linter (note: Linter) reports (47 passages, no note) the same finding in global scope (note: Scope) for a local (27 passages, no note) title that an entity of another type (3 passages, no note) carries in the published model (note: Model). A twin (note: Twin resources) folded into its note (note: Note) is no homonym (note: Homonym): the twin resources are reconciled (3 passages, no note) before the dictionary is built, and the titles and aliases of the folded twin name the entity it became, once.
Check (note: Check) I-TERM-HOMONYM, severity (note: Severity) info (8 passages, no note) by default, documented with a before and after example in the check pages. The severity can be overridden per project (note: Project) or per repository (note: Knowledge repository).