Core meaning
- Capability purpose
- Surface useful vocabulary discoveries while keeping inference separate from owner-confirmed meaning.
- Meaningful outcome
- A candidate, recognized occurrence, or ignored result is recorded with source evidence and confidence.
- Boundaries — includes
- source-scoped detection, normalization, known-term matching, candidate evidence, confidence, ignore handling, and audit.
- Boundaries — excludes
- confirming a glossary term, changing source content, globally resolving identity, and presenting a specific highlight UI.
- Terms and concepts
- A candidate may map to a
DefinedTerm; recognition is an evidence claim, not identity truth.
Interaction Contract MLEs
Detect term candidates
- Actor
- An authorized owner or enrichment process.
- Command / intent
- Examine a governed source for terms relevant to the personal glossary.
- Current state
- Source content and the applicable glossary/ignore context are available.
- Policies / invariants
- Source provenance is retained; ignored candidates are respected; known aliases resolve without creating duplicates; new candidates remain unconfirmed; uncertain identity matches are not silently globalized.
- Transition
- Analyze the source, normalize candidates, compare known terms and exclusions, and record candidates or occurrences with evidence.
- Result
- Evidence-linked candidates/occurrences or a reasoned no-result/failure.
- Events / effects
- A new candidate may trigger the governed proposal capability.
- Unknowns
- Detection model and threshold are realization choices.
Rules and defaults
Rules / invariants
- Detection must not silently create accepted glossary meaning.
Recommended defaults
- Deduplicate against active terms and aliases before proposing a new term.
Unknown / unresolved
- Public implementation evidence is absent.
Statement provenance
| Statement | Semantic class | Evidence status | Source / note |
|---|
| Detected names remain candidates until governed confirmation. | rule/invariant | sourced | K03. |