One English word names six unrelated ideas. The Concepts editor keeps them apart.

Most terminology projects start in a spreadsheet. Someone adds every word that looks technical. A year later the file holds four thousand rows, and nobody can say which rows were ever agreed.

A termbase records the ideas your organisation cannot afford to name inconsistently. If you organise it by word, three of your entries can name the same idea and you will not notice. If you organise it by concept, you see the overlap when you create the third one.

The TextUnited Concepts editor is built on that second model.

Concept #232322 carries three labels in three languages

Let’s start with a fully built entry.

The concept is numbered #232322 and named AeroCord. It carries three terms: power cord in English (US), Netzkabel in Swiss German and kabel sieciowy in Polish. The concept has its own number and its own name, so none of the three labels stands in for the idea itself.

That arrangement changes who is upstream. In a termbase organised around a source term, the English word becomes the concept, and every other language holds a translation of an English word rather than a name for an idea. Give the concept its own identity and the Polish label sits at the same level as the English one.

Around that identity sit the fields that make the idea unambiguous.

The Domain field reads Electric & Electronics and the Class field reads Object/component. Domain records where in the business the idea belongs. Class records what kind of thing it is.

The definition sits at concept level: a US power cord typically uses the North American plug type according to the NEMA standard with a mains voltage of 120 volts. The termbase holds that definition once. It does not repeat it per language, and it does not attach it to whichever term someone opened first.

The description holds the guidance a definition should not carry: pay attention to both ends of the cable, the wall plug and the device connector.

The note records produced at Ohio plant. The tag records deliverable. Neither field belongs to the meaning, and the people using the entry need both.

The image is a photograph of the cable, with the NEMA plug at one end and the device connector at the other.

That last field deserves its own paragraph, because terminology theory does not ask for it. Concept boundaries in industrial content are frequently physical, and you can show a physical boundary faster than you can describe one. You can write three careful definitions separating a box, a unit and a switch box, and a translator can still hesitate. Three photographs show the difference. For enclosures, connectors, fittings and assemblies, the photograph is often the most precise boundary marker available, and the fastest one to read at the moment someone needs it.

Below the shared fields sit the terms themselves. Each of the three carries a modification time, a named author and a status. All three currently read Accepted, and that status has a specific consequence, described further down this page.

Relations connect one concept to another concept

The Relations panel on AeroCord holds one entry: PARTS, part of, fuente de alimentación.

AeroCord is recorded as part of another concept, and you can follow that link. You add relations through + Add relation, and the editor groups them by kind. Relations connect concepts to concepts. Terms never link to terms, and that restriction keeps the model from collapsing into a word list.

Most termbases skip this step, and it is the step that changes what the termbase is. A set of well-defined entries answers lookups. A set of entries that records how they fit together does four further things.

It shows overlaps that definitions miss. Two concepts can each be defined correctly and still overlap, or leave a gap between them. You see that when they sit next to each other in a structure. Partitive relations such as part of expose it most often.

It records which pairs get confused. The most useful relation in industrial terminology is not hierarchical. It is the statement that two concepts are easy to mistake for each other.

It lets navigation follow meaning. You can move from a cable to the power supply it belongs to. Searching for a string does not do that, and the graph is how someone explores a domain they do not already know.

It disambiguates where strings cannot. This termbase holds two separate concepts whose English label is power supply. One carries Netzteil and fuente de alimentación. The other carries Speisung. AeroCord is recorded as part of the first one. A cross-reference pointing at the word power supply leaves the reader choosing between two ideas. A relation pointing at a concept does not.

It supports inheritance and validation. Domain and Class can default down a hierarchy, which removes work from classification. Structure also makes some errors cheap to catch: an object whose broader concept is a process is almost always a mistake, and software can flag that without anyone writing a rule.

Each relation has a type. A partitive relation and a hierarchical relation mean different things, imply different cardinality and support different reasoning. Typing them is what makes the four points above possible, and typed relations survive a TBX export with their meaning intact.

Six unrelated ideas share the English word power

Search the termbase for power. The results come back grouped by language.

• English (UK): power, powerful, empowerment, sound power level
• English (US): extraction power, IEC power connector, power box, power connection, power cord, power input, power supply, power supply filter, power switch box, power unit
• German (Swiss): Absaugleistung, Kaltgerätestecker, Leistungsaufnahme, Netzfilter, Netzkabel, Netzteil, P-Box, Powerbox, Powerschaltbox, Schallleistungspegel, Speisung, Stromanschluss
• German (Germany): kraftvollen, Vertretungsbefugnis
• Danish: styrke · Dutch: voedingskast · Japanese: 電源接続 · Odia: aangoo · Polish: kabel sieciowy · Spanish: fuente de alimentación

One English string sits above an acoustic measurement (sound power level, Schallleistungspegel), a suction capacity (extraction power, Absaugleistung), an electrical draw (power input, Leistungsaufnahme), a physical cable (Netzkabel, kabel sieciowy), the legal authority to represent someone (Vertretungsbefugnis) and abstract strength (styrke, aangoo). These ideas have nothing to do with each other. A word-oriented glossary files them as neighbours, and a translator working under deadline picks one of them.

Note also that IEC power connector maps to Kaltgerätestecker. The German word contains no part of the English one. String matching never pairs them. Concept matching returns one answer.

Select a concept and the editor highlights every term it carries, each in its own language branch. The tree renders one set of concepts from each language's point of view.

The form asks for the definition first

In the Create New Concept dialog, the Definition field comes first, prompting you to enter the core definition of this concept. Domain, Class, Description, Note and Tags sit alongside it. Terms come last, at the bottom, added one row at a time with a language, the term text and a status.

The layout asks you to say what the idea is before you say what it is called.

The filters catch duplicates

The left panel looks like a browser. It also works as a gate.

You can filter by Language, Domain, Class, Status, Tags and Import, and by creation date and modification date ranges. Those filters find entries. They also run the check that belongs before every new entry: search across all languages and all statuses first, because the concept you are about to create often exists already under a label you did not think of.

The date ranges carry the governance work. You can list everything created in the 2016 import, everything untouched since the last product generation, or everything added since the last ISO audit. A review cadence only works if you can slice the termbase by age.

The domain folders in the tree make classification visible. Electric & Electronics holds three concepts. No specific domain holds thirty-eight. Those thirty-eight concepts still need a domain, and nobody sees that until the structure shows it.

The Quick Concept Editor shows every concept in every language at once

The concept view is where you work on one idea in full. The Quick Concept Editor is where you read the termbase as a whole, and that view only works if concepts are the underlying object.

Rows are concepts, eighteen of them here. Columns are languages, twelve in this account, and the grid scrolls horizontally when the language set gets wide. Every cell either holds a term or shows + Add, and cells that already hold a term show + Add as well, because a concept can carry several labels in one language.

Colour carries the status. Blue marks accepted, green marks enforced and red marks rejected, across the whole termbase without opening a single entry.

This account shows a lot of red and very little green. Two terms are enforced, sound power level and Schallleistungspegel, plus the Danish styrke. Everything else is rejected or waiting. A vocabulary of enclosures, connectors and cables has no verdict attached, and enforcement is what binds the translation engine, so almost none of this terminology currently constrains the output.

Look at the row holding styrke. The Danish term is green. The English power and the Odia aangoo on the same row are red. Status belongs to a label in one language, not to the concept, because the question "is this the correct word here" is answered per language. The concept holds the meaning and each language decides its own labels.

Five further things follow from the layout.

The curated concept looks different from the rest. Row one is AeroCord, with Domain reading Electric & Electronics and three blue terms in English, Swiss German and Polish. Every other row shows a dash in the Domain column. One curated entry sits against seventeen unclassified ones, and you can see which is which without opening any of them.

Coverage gaps become readable. The question "which languages is this idea missing" turns into a row you read rather than a report you run. Swiss German covers most of the cluster. Japanese covers one concept, 電源接続 for power connection. Danish, Dutch, Odia, Polish and Spanish cover one each. Those gaps are not a translation backlog. They are decisions nobody has made, and whoever next translates a manual under deadline will make each one alone.

Homographs separate cleanly. Power supply appears on two different rows in English (US). One row carries Netzteil and fuente de alimentación. The other carries Speisung. One English label covers two distinct ideas, the physical unit and the feed, and the termbase holds them as two concepts. A word list shows two identical rows and leaves you guessing.

Confusable neighbours line up. Read down the Swiss German column: power box maps to Powerbox, power unit maps to P-Box, power switch box maps to Powerschaltbox. Three near-identical labels stand for three supposedly different things, and power box also carries the Dutch voedingskast, which names a cabinet rather than a box. The labels alone do not tell you whether these are three concepts with a boundary problem or one concept with three names.

That cluster is what definitions, images and relations exist for. Whatever separates a power box from a power unit and a power switch box, you can photograph the difference. Once someone resolves it, the fact that the three entries are worth checking against each other gets recorded once instead of rediscovered by everyone who opens the termbase later. The grid does not resolve the cluster. It shows you which three entries to open first.

Orphan concepts surface. One row holds Powerbank in German (Austria) and powerbank in Swedish. Another holds powered and driven. Both English columns are empty. A concept with no source-language anchor is hard to govern and easy to duplicate, because the next person looking for that idea searches in English, finds nothing and creates it again.

The Domain column carries a dropdown on every row. Row checkboxes plus per-row dropdowns turn classification into a batch operation instead of seventeen trips through the concept view. Bulk status changes and post-import cleanup work the same way. Inline + Add means a reviewer working down a language column can enter an equivalent without leaving the grid.

TextUnited imports and exports CSV and TBX

Almost nobody starts from zero. A technical writer has been maintaining a spreadsheet. The LSP you are leaving has a TBX. A client sent a glossary as a condition of the contract. So the first question about any termbase tool is what it accepts.

Import takes CSV, TBX and TBXM. TBX is the ISO exchange standard for terminology, so a termbase can move between systems with its structure intact. CSV is the format your subject-matter experts already have. The import runs as a stepped flow, so you see the contents before you commit them.

Two things about imports are worth saying plainly.

First, an import is where technical debt enters the termbase. Entries from 2016 with no definitions. Thirty-eight concepts with no domain. Orphan concepts with no source-language anchor. That is what an inherited termbase contains, and no import wizard cleans it for you. The platform makes the debt visible and cheap to work through: the Import filter isolates exactly what arrived in one batch, and the Quick Concept Editor classifies, restatuses and fills entries in bulk. Import, isolate, clean, approve, and you have an onboarding path you can finish.

Second, an import proposes entries. Everything that lands should land as a candidate.

Export uses the same formats as import, CSV or TBX, with filters for Domains and Languages. An empty selection means all.

This is a deliberate position. Your terminology is your intellectual property. It records decisions your engineers, your legal team and your subject-matter experts made about how your products are named, and it is often the most valuable linguistic asset a manufacturer owns. It should not sit inside one vendor's system. If you leave TextUnited tomorrow, you leave with a standards-compliant TBX holding everything you brought in and everything you built here.

The filters also make export useful day to day.

• Filter by language, and hand a new Polish supplier the Polish column and its English source, and nothing else.
• Filter by domain, and send the regulated product line's terminology to the notified body without shipping the rest of your termbase.
• Select preferred terms only, and export one agreed label per concept per language instead of every synonym, variant and rejected proposal. A machine translation glossary needs that shape. So does a QA check, and so does an authoring checker. Your full termbase records the discussion. A preferred-terms export carries the decision.

Inside TextUnited you do not need the export for this. The termbase is already wired into translation, and the API and the MCP server put it in front of whatever else you are orchestrating. The export serves everywhere else: the agency you work with, the CMS vendor, the notified body, and the system you might move to one day.

A term entry records one label in one language

The concept carries the meaning. The term carries everything that belongs to one label in one language.

Language. EN-US and EN-UK are separate. So are German (Germany), German (Austria) and German (Swiss). For industrial clients shipping into DACH, that granularity records a fact worth keeping: Netzkabel is the Swiss term, and Germany may settle on a different word.

Status. Rejected, Accepted or Enforced. This field decides whether people can rely on the termbase, and it produces the colour in the grid.

Abbreviation. The short form that appears in UI strings, drawings and parts lists, recorded against the term it shortens instead of standing as a separate entry.

Notes. Scope and caveats that apply to this label in this language. The core definition sits one level up.

Sample. Real usage in context. A definition says what the concept is. A sample shows how the term behaves in a sentence. Samples also test distinctness: if the samples for two concepts are interchangeable, you have one concept.

Tags. Classification that cuts across the domain tree, such as product line, regulatory scope, customer or review batch.

The audit trail. Author, created, modified, and separately, who approved the term. AeroCord's terms show named authors and modification times measured in minutes, against a concept created in October 2016. That entry is old and someone is still curating it.

Status separates a correct term from a compulsory one

A term sitting in the database is not the same as a term cleared for use, and a term cleared for use is not the same as a term that binds. The status ladder records both distinctions.

Rejected. Someone considered the term and turned it down. The entry stays visible, so nobody proposes it again next year.

Accepted. A subject-matter expert has confirmed that the concept is distinct, necessary and correctly defined, and that this label is right for this language. The term is correct and usable, and it does not yet bind anything.

Enforced. The term binds across the whole account: every project, every user, every automated job.

That middle rung does more work than it appears to. Most termbases have one gate, which forces a choice. Either approval is cheap and the database fills with entries nobody has verified, or approval is expensive and nothing gets approved. Two gates separate two questions that were always separate. The first question is whether the term is correct, and a person who knows the subject answers it. The second question is whether the term should bind every project in the account, and whoever owns that commitment answers it.

AeroCord shows the ladder in use. The concept has a definition, a class, an image, a relation and three languages, and all three terms read Accepted. Someone validated the entry and deliberately did not impose it on every project yet.

Without a status ladder, a termbase turns back into a word dump. Writers and translators treat everything they find as binding, including half-formed proposals and entries a subject-matter expert killed years ago. With the ladder, the pipeline can stay open and low-friction, because candidates are visibly candidates, and the Quick Concept Editor renders all three rungs in colour.

Status is also what makes preferred terms only a meaningful export. A termbase that exports its whole contents regardless of status tells you what it holds. One that exports the verdict tells you what your organisation agreed.

Terminology also ages. Products get renamed, meanings shift, and a term that was correct three product generations ago goes stale. The modification-date filter finds those entries, which is why the editor records when and by whom rather than only what.

Enforced terminology binds automated translation

In TextUnited, an enforced term constrains automated translation. The engine does not consult it as a reference and nobody counts violations in a report afterwards. Once a term reaches Enforced, the machine uses that term on that job and on every job after it, across the account.

The top of the ladder is not called final, published or approved. It is called Enforced, and the name states the consequence, so the person setting it knows they are changing what comes out of the pipeline.

The top rung is therefore not administrative metadata. It is a switch between recorded and binding, and promoting a term to it changes production output immediately. That is why it sits above Accepted rather than beside it, and why the editor records it as a named, dated, separate event.

It also explains why the concept model is not a preference. Enforcement is only safe if the match happens at the level of meaning.

Consider what a string-level glossary does with the termbase above. Tell it to enforce Netzkabel wherever it sees power, and it will rewrite the acoustics content, the extraction specifications and every general use of the word. The concepts under power require different answers: Schallleistungspegel in one context, Leistungsaufnahme in another, Kaltgerätestecker in a third, and no substitution at all in a fourth. Only a concept-oriented termbase can carry that instruction, because only a concept-oriented termbase knows which idea is on the table.

This is what the supervised model looks like in production. A subject-matter expert makes one judgement, about this concept, this label and this language, and that judgement constrains machine output at volume afterwards without the expert being present. The termbase is how human expertise reaches past human throughput. Everything else in the Concepts editor exists to make sure the judgements being enforced are the right ones.

The assistant drafts definitions and usage examples

The assistant opens alongside the concept you are working on, and the context card at the top states its scope: you are viewing this concept, in these languages.

The assistant is scoped to your termbase and to the concept in front of you. It searches your concepts, checks translation coverage across your language set, suggests terminology and flags where consistency is slipping.

Its starter actions cover the parts of the work that are slow and mechanical, and that people therefore skip.

Explain the concept. Everyone inherits a termbase eventually, with entries created years ago by someone who left, carrying no definition and no note about the decision. A first-pass reading grounded in how the term appears across your content shortens the research, and post-import triage is where you will use it most.

Suggest synonyms. This runs the uniqueness test in reverse. Ask for the plausible alternative labels for a concept, then search each one. Anything that returns an existing entry is a merge candidate, and the fix for two labels naming one idea is a second term under the same concept rather than a second concept.

Write a clear definition. People skip definitions because writing one is slow. A usable definition names the broader category the concept belongs to, then the feature that separates it from its siblings: a [broader thing] that [distinguishing feature]. No circularity, no or holding two concepts together in one sentence, and scoped to how your industry uses the term rather than to the fullest abstract sense. The assistant drafts into that shape, which turns write a definition into check this definition.

Give usage examples. Drafts for the Sample field, taken from your own content rather than invented.

The assistant does not decide. It drafts, researches and surfaces gaps. A subject-matter expert still confirms that the concept is distinct, necessary and correctly defined, and a named owner still sets the status that binds the engine. That division is the same supervised model the rest of the platform uses. The machine handles volume and first drafts. The person holds the authority, and the audit trail records who held it.

The MCP server opens the same termbase to assistants and agents

The Concepts functionality is not confined to the terminology manager. The TextUnited MCP server exposes it, so an MCP-capable assistant or agent can reach the same concepts, the same languages and the same statuses that the web app shows.

That matters because terminology questions rarely arise inside the terminology tool. They arise while someone drafts a manual, reviews a supplier's copy, writes release notes or builds a pipeline. Making the termbase reachable from those places removes the step people skip, which is switching applications to check.

Ordinary questions become one query instead of a detour. Is there an enforced term for this concept in Polish? Which languages is this idea missing? Does this phrase already exist under another label? What is this concept part of? A pipeline can also check coverage before a job runs rather than finding the gap in a QA report afterwards.

The governance boundary holds. An agent can search, retrieve, compare and propose. Deciding what the engine is bound by remains a human act with a name attached, for the reason the previous section gives. The MCP server widens who can ask the termbase questions. It does not widen who can change what the termbase enforces.

The web app supports curation. The API supports integration. The MCP server supports assistants and agents. All three reach one termbase.

A concept moves from candidate to enforced in six steps

1. Import or propose. Bring in an existing CSV or TBX, or let anyone nominate a concept where confusion showed up. Either route lands the entry as a candidate, tagged to its import batch.
2. Research. Search across all languages and all statuses to confirm that the concept does not already exist under another label. Use the assistant to gather usage and draft the definition.
3. Structure. Assign domain and class, write the definition, attach an image if the boundary is physical, and record how the concept relates to its neighbours.
4. Accept. A subject-matter expert confirms that the concept is distinct, necessary and correctly defined, and that this label is right for this language.
5. Enforce. The term goes from correct to compulsory across the account. Enforced terminology binds automated translation from that point, and it reaches writers, partners and connected tools through the API, the MCP server or a preferred-terms export.
6. Review. Filter by modification date to find what is ageing. Open the Quick Concept Editor to find what is missing.

Start with the idea

The test of a termbase is not how many rows it holds. The test is whether the right word comes out the other end, whether a translator working at 4pm against a deadline relies on what they find, and whether the engine running overnight is bound by it.

That reliance comes from curation: a tight set of concepts that are unique, cleanly bounded against their neighbours, expensive to get wrong, settled enough to define, and frequent enough to justify the upkeep. Everything else is noise, and noise is why people stop checking.

Concept #232322 shows the finished state. It has a name of its own, a domain, a class, a definition, a photograph, a place in a structure, and three accepted labels in three languages, one promotion away from binding every job in the account. Nobody needs computer in the termbase. They need the concepts their industry argues about, built to that standard: defined once, agreed by the people who know, enforced automatically once someone decides they should be, and portable enough to remain theirs.

The Concepts editor, Quick Concept Editor and Terminology Assistant are part of the TextUnited terminology manager. Concepts are available in the web app, through the TextUnited API, and through the TextUnited MCP server for use with MCP-capable assistants and agents. Terminology imports and exports support CSV and TBX.