River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Technical Glossary and Terminology Template

Two documents and three sheets that locate every occurrence of every rejected term across your docs, UI strings and support content.

Free download  ·  No account needed

Inconsistency Report

[Product], [n] concepts tracked

Against [n] doc pages, [n] UI strings, [n] support macros

The register, in four counts

Concepts checked
Concepts with more than one term in use
Total occurrences logged
Non-approved occurrences: the remediation debt

The number a glossary entry alone never produces

Not which term won. How many locations, named individually, still say the other one, split by documentation page, UI string and support macro, each with an owner.

Ranked by remediation size

Every inconsistent concept, ordered by non-approved occurrences rather than by how untidy it looks, because that ordering is the work queue.

The override column

Where the approved term is not simply the majority, the reason stated on the glossary entry itself.

The remediation tracker

One row per location. A state column: not started, in progress, blocked, done.

Terminology drift shows up as the same entity called an account on forty pages and an organization on twenty, and every writer who reaches for a glossary reaches for the same fix: name the correct word and stop. Redocly's own guidance gets further than most, converting an audit into a one-page glossary and enforcing forbidden variants going forward with CI lint rules. Contextive's glossary files take the same shape: terms, definitions, and a context they apply within.

Neither one counts what already exists. A CI rule catches the next pull request; it does nothing about the ninety-eight existing pages already carrying the rejected word. This pack searches three surfaces instead of one: documentation pages, UI strings, and support macros or canned replies, because a UI label often predates the documentation team's preferred term entirely. On the worked corpus, Fenwick Ledger, an invented billing product, 64 concepts were checked and 23 came back with more than one term genuinely in use, 35.9 percent.

Those 23 concepts carry 2,180 occurrences that are not the approved word, spread across 640 distinct locations. That 640 becomes the Remediation Tracker: one row per location, a surface, an owner, and a state, so fixing the largest concept first is a queue somebody works through rather than a paragraph in a glossary nobody reopens. Scoring a page's prose against the finished glossary is a separate pass, the copyedit and style pack; whether the facts on an existing page are still true, rather than which word names them, is the documentation audit.

64 concepts checked, 23 with competing terms, 2,180 occurrences that are not the approved word

The Term Register, the Inconsistency Report and the Remediation Tracker.

Term Register

Fenwick Ledger, an invented billing product. 187 doc pages, 640 UI strings, 1,340 support macros searched. Concept: the top-level paying entity, 360 occurrences across 115 locations.

TermDoc pagesUI stringsSupport macrosTotal occ.Approved
account41 / 15422 / 528 / 10216Yes, 60.0%
organization19 / 715 / 193 / 898No
tenant2 / 69 / 241 / 434No
workspace1 / 32 / 62 / 312No

Columns read locations / occurrences. 41 pages use "account" and 19 use "organization" for the same concept, so a reader moving between two pages sees two different words for the entity they are looking at.

Inconsistency Report

23 of 64 concepts, 35.9%, have more than one term genuinely in use. Ranked by remediation occurrences, the actual work queue.

RankConceptTotal occ.ApprovedShareRemediation occ.Locations
1The top-level paying entity360account60.0%14444
2The recurring charge on a schedule395subscription79.0%8320
3The refund approver role41billing admin58.5%179
4The exported record of transactions61statement91.8%53
5-2319 further conceptssee register1,931564
Total, 23 concepts2,180640

The refund approver role ranks third by remediation size despite being the smallest concept by total occurrences, because a smaller concept with a weaker majority (58.5%) carries more relative debt than the second-ranked one at 79.0%.

Remediation Tracker

One row per distinct location, never one row per concept. 640 rows total; 8 shown.

LocationSurfaceCurrentApprovedOwnerState
/accounts/overviewDoc pageorganizationaccountDocsNot started
settings.tenant_picker.tooltipUI stringtenantaccountEngineeringNot started
macro-142-org-transferSupport macroorganizationaccountSupportDone
/billing/plans-vs-subscriptionsDoc pageplansubscriptionDocsIn progress
roles.account_owner.refund_permission.labelUI stringaccount ownerbilling adminEngineeringBlocked
/reports/ledger-exportDoc pageledger exportstatementDocsNot started

Two rows are blocked, not started: one on an API endpoint alias that needs engineering sign-off before the doc page can change, one on a role-model decision that needs recording first. Blocked is a different state from not started, and the tracker keeps them apart.

What's in the pack

01

Term Register

Every candidate term for every concept, counted by location and by occurrence across documentation, UI strings and support content.

02

Inconsistency Report

Concepts with more than one term in use, ranked by remediation size rather than by how untidy each one looks.

03

Remediation Tracker

One row per distinct location that needs to change, with a surface, an owner and a state, never a grouped row per concept.

04

Glossary

The approved term and every rejected synonym per concept, with the override reason stated whenever the choice was not the plain majority.

05

Usage Guidance

Five checks for whoever writes the next page, string or macro, short enough to actually reread before typing.

06

Cross-surface occurrence counting

UI strings and support macros searched with the same rigor as documentation, because drift is often worse where review is lighter.

07

A space rule every prompt reads first

A term is not standardized until every occurrence is located, not merely until a glossary entry declares a winner.

How to use it

  1. 1

    Open it in River, or download it

    Open the pack and the agent searches your own docs, UI strings and support content before naming a single term approved, or download the blank sheets and docs.

  2. 2

    Send your corpus and your candidate terms

    Your documentation set, a UI string export, a sample of support macros, and any concepts you already suspect have competing terms.

  3. 3

    Read the occurrence counts before any decision

    Every candidate term is counted by location and occurrence on all three surfaces, so the approved term is chosen from evidence.

  4. 4

    Work the remediation queue

    Ranked by remediation size, with one tracked row per location, a surface, an owner, and a state you can total for real completion.

Frequently asked questions

Is this template free?

Yes. Download the whole pack as Word documents and CSV sheets, no signup and no credit card. Edit with AI is a separate, optional path for anyone who wants the agent to search their own corpus. Nothing happens until you send it something to check.

We already have a glossary. What does this add?

Most glossaries name the approved term and stop, which fixes nothing already written. This searches your documentation, UI strings and support content for every occurrence of every rejected synonym, then builds a tracked row for each location, so an existing glossary gets a remediation queue behind it instead of just a policy statement.

Do we need access to our UI strings and support macros, or is the documentation enough?

Documentation alone misses most of the real drift. A UI label often predates the documentation team's preferred term, and a support macro can preserve a term from two product renames ago. On the worked example, over a third of one concept's occurrences lived in UI strings and support macros rather than in prose.

How does it decide which term wins when we disagree internally?

It defaults to the term with the most occurrences, since that is what most of the corpus and most readers are already trained on, and states the exact share. You can override it, but the override needs a stated reason on the glossary entry, such as a collision with an already-approved term for something else.

Will it rewrite our pages and strings for us?

No. It locates every occurrence and builds the tracked row for each one, with a surface and a suggested owner, but making the actual edit stays with whoever owns that documentation page, UI string or support macro. The tracker is a work queue, not an automated find-and-replace.

What format are the downloaded files?

Two Word documents and three CSV sheets, zipped. The documents open in Word, Pages and Google Docs; the sheets open in Excel, Numbers and Sheets. Add a PDF query string to the download if you want to circulate the glossary rather than fill it in.

How is this different from the style guide in the docs-as-code pack?

Different layer. A house style guide covers grammar, tone and sentence-level rules that a linter can enforce going forward, which is what the docs as code workflow pack checks. This is about which noun means which concept, and it reaches backward into everything already written, not just the next pull request.

Find out how many pages are still using the word you rejected

Download the blank Term Register, Inconsistency Report and Remediation Tracker as Word and CSV, or have River search your own docs, UI strings and support content.

Edit with AI