Terminology Audit¶
Diagnostic review — instantiates Semantic Drift Monitoring
Cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift.
A Terminology Audit is a bounded, point-in-time cross-check: it lays a term's definition of record beside a deliberately gathered sample of how the term is actually being used now, and renders a verdict on whether — and in what shape — the two have come apart. It has a start and an end. You name the term, freeze its baseline meaning, pull a usage sample, and classify the result against a fixed typology: still aligned, narrowed, widened, split into senses, shifted in valence, or moved to a new domain. That classified verdict is its entire output. The audit does not watch a live stream, and it does not rewrite the definition — it is the diagnosis that tells the rest of the loop whether either of those is warranted.
Example¶
A national photograph archive runs an audit on the collection-status word "processed." The processing manual (the definition of record) says a collection is "processed" only when every folder has been described to the item level. But archivists have grumbled that the word no longer means what the manual says. The auditor freezes the manual's definition as the baseline, then pulls a bounded sample — 200 finding aids published in the last three years — and reads how "processed" is used in each. The pattern is unmistakable: in more than half the recent aids, "processed" is applied to minimally processed collections described only to the box level. The verdict is a single classified finding: the term has widened, absorbing a lighter standard the manual never sanctioned. The audit ends there. It does not decide whether to accept the wider meaning or restore the narrow one — it hands the widening verdict to whoever owns the definition, having turned a vague staff complaint into a documented, sampled diagnosis.
How it works¶
- Name the unit precisely. One term, heading, or code — not a topic area. Ambiguity about what is being audited sinks the result.
- Freeze the baseline. Copy the definition of record verbatim, with its source and date, so the comparison has a fixed anchor rather than a remembered one.
- Draw a bounded usage sample. A finite, defensible slice of current use — a set of documents, records, or tickets — read closely. This is a snapshot, deliberately not a standing observation feed.
- Classify against a typology. Place sample beside baseline and assign the drift a shape: aligned / narrowed / widened / sense-split / valence-shift / domain-shift. The fixed typology is what stops "it feels different" from becoming the finding.
- Record the verdict with its evidence. The classified call plus the sampled instances that support it, ready to hand off.
Tuning parameters¶
- Sample size and scope — more instances and more sources give a more trustworthy call, at more reading effort; a thin sample risks over- or under-calling drift.
- Typology granularity — a coarse aligned/drifted split is fast; a fine-grained shape typology guides the eventual fix but takes more judgment to apply consistently.
- Baseline authority — which definition counts as "of record" (the manual, the founding memo, the last ratified glossary); a weakly authoritative baseline makes every divergence arguable.
- Cadence — one-off versus a repeated audit on a schedule; repetition catches slow drift but spends reviewer time on terms that may be stable.
When it helps, and when it misleads¶
Its strength is conversion: it turns a diffuse sense that "people don't mean the same thing anymore" into a sampled, classified verdict that a definition owner can actually act on — and it is cheap and repeatable enough to run on many terms. Its characteristic failure is the recency illusion — mistaking a usage the auditor has only just noticed for one that is genuinely new[1] — which a thin or convenience sample amplifies into a false drift call. The verdict is also only as trustworthy as the baseline's authority: audit against a definition nobody ratified and every divergence is contestable. The guarding discipline is to require a real sample rather than a handful of vivid instances, and to treat a single audit's verdict as provisional — a snapshot, not a trend — until a wider look confirms it.
How it implements the components¶
term_or_symbol_under_watch— the audit's first act is naming the exact term, heading, or code whose meaning is in question, at a resolution tight enough to sample.baseline_meaning_snapshot— it freezes the definition of record, with source and date, as the fixed reference the current sample is read against.meaning_scope_change_assessment— its verdict is the classification of the drift's shape (narrowed, widened, split, valence, domain), against a fixed typology.
It does not run a standing usage_observation_window or emit an emerging_meaning_signal — that continuous sensing is Corpus / Usage Monitoring, the nearest twin: monitoring watches a live stream and never stops, while the audit is a one-shot cross-check with a verdict. Nor does it own the definition_update_rule that acts on the verdict — that belongs to Glossary Update Workflow.
Related¶
- Instantiates: Semantic Drift Monitoring — the audit is the diagnostic step that decides whether drift is real and what shape it has.
- Consumes: Corpus / Usage Monitoring — when a monitored corpus exists, the audit can draw its usage sample from it rather than gathering one by hand.
- Sibling mechanisms: Corpus / Usage Monitoring · Glossary Update Workflow · Definition Change Log · Semantic Versioning Policy · Policy Language Review · Stakeholder Interpretation Check · Usage Exception Registry
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Terminology Audit operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift.
Independent corroboration: The frozen evidence defines Terminology Audit as 'Cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift', so its operative form is Assessment, Review & Assurance.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Linguistics & Semiotics
Origin pattern: Single lineage
Present-day reach: Universal
Rationale: The defining operation is: Cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift. In the linguistics_semiotics lineage, that operation is specifically evidenced by authoritative or primary work that grounds concept definitions, term assignment, consistency, equivalence, and cross-system terminology alignment. This makes linguistics_semiotics the best historical origin, while the retained alternates document contributing methods and later applications rather than being mistaken for coequal origins.
Related originating lineages:
- Communication & Media Studies — Communication and media research supplies a parallel or contributing lineage for the mechanism's defining operation: cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift.
- Library & Information Science — Library and information-science stewardship supplies a parallel or contributing lineage for the mechanism's defining operation: cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift.
Review resolution: The blind reviewers disagree on primary lineage (library_information_science versus linguistics_semiotics), so I adjudicated the mechanism rather than inheriting either label. The defining operation is: Cross-checks a term's official definition against a sampled snapshot of how it is actually used, and classifies the shape of any drift. In the linguistics_semiotics lineage, that operation is specifically evidenced by authoritative or primary work that grounds concept definitions, term assignment, consistency, equivalence, and cross-system terminology alignment. This makes linguistics_semiotics the best historical origin, while the retained alternates document contributing methods and later applications rather than being mistaken for coequal origins. The cited ISO 704 Terminology work — Principles and methods directly supports the mechanism-specific operation and its disciplinary lineage. I retain all independently explained historical alternates without a numeric cap. origin_mode=single_lineage records how the mechanism arose; domain_reach=universal separately records how broadly it can now be applied.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
References¶
[1] Zwicky, A. "Just between Dr. Language and I." Language Log (2005). Defines the Recency Illusion as mistaking a recently noticed usage for a genuinely recent one. registry ↩