On this page
Three categories, three usersCAT tools: the translator's workbenchTMS: plumbing for continuous contentMachine translation: fast drafts, with conditionsOther tools you will hear aboutBuyer's checklistShort answersA procurement manager gets three software demos in one week. The first shows a translator's editor with matches and term suggestions. The second shows a dashboard that pulls strings from a code repository and routes them to reviewers. The third shows a box where English goes in and Korean comes out instantly. All three are called "translation software". They solve different problems, are used by different people, and are bought with different budgets. Before asking what a CAT tool, a TMS or an MT engine costs, it helps to know which problem each one solves.
This post sorts translation software into its main categories, from a buyer's point of view: what each does, who sits in front of it, what you get out of it, and what to check before signing. For the wider tool landscape for small teams, see free vs paid translation tools for small businesses.
Three categories, three users
| CAT tool | Translation management system (TMS) | Machine translation (MT) | |
|---|---|---|---|
| What it does | Helps a translator work: editor, translation memory, termbase, QA checks | Moves content: connects sources, assigns work, tracks status, returns files | Produces a draft translation automatically |
| Who uses it | Translators and revisers | Localization managers, project managers, developers | Anyone; or another system through an API |
| What you own afterwards | Translation memory and termbase | Workflow history, memory, glossary, reports | Nothing persistent unless you add glossaries or custom models |
| Main buying question | Does your vendor already use one? | How often does your content change, and from how many places? | Where may your data go, and who checks the output? |
Many products now blend categories: a TMS with a built-in editor and MT, or a CAT tool with a cloud workflow. Still, ask which of the three jobs you actually need done. That decides the budget line and who evaluates the product.
CAT tools: the translator's workbench
Computer-assisted translation tools split text into segments and show the translator, for each one, any earlier translation of the same or a similar sentence from the translation memory. Microsoft's localization guidance describes a translation memory as a database of source segments paired with their translations, and explains that partial "fuzzy" matches get a percentage score depending on how much the sentence differs. The translator decides whether to accept, edit or redo each match. Termbases do the same for individual terms.
For a buyer, the point is not the editor but the by-products. Your memory and termbase are the reason the second edition of a manual costs less than the first, and the reason "torque limiter" is translated the same way in every chapter. The standard exchange formats matter here: TMX for translation memories, XLIFF (an OASIS standard) for bilingual files, and TBX for terminology. Ask any vendor to deliver your memory in TMX on request. More on that in what translation memory is and why you should own it, and for how CAT tools are used on patents, CAT tools in patent translation.
TMS: plumbing for continuous content
A translation management system is workflow software. It connects to where your content lives (a code repository, a CMS, a help center, a design tool), detects new or changed text, sends it through pre-translation and human steps, and returns it. It tracks who is doing what, by when, and what it cost.
You need one when content changes continuously and comes from several sources, which describes most software companies and online retailers. You probably do not need one if you send a few documents a quarter. The usual failure is buying a TMS and connecting only the app, while marketing emails and legal pages still travel by email attachment. For a step-by-step setup, see automating a localization pipeline.
Machine translation: fast drafts, with conditions
MT engines are available as consumer apps, business subscriptions and APIs, and are often built into TMS and CAT tools. For Korean, quality has improved a lot, but the familiar weaknesses remain: level of formality chosen inconsistently, subjects that English states and Korean omits handled unpredictably, and terms translated differently from one sentence to the next.
Two features matter most to a buyer. First, glossaries: Google's Cloud Translation Advanced, for example, lets you define a glossary, a custom dictionary of terms, either for one language pair or as equivalent term sets across several languages, so that product names and key terms come out as you specify. Second, data handling: DeepL's business security page states that texts are never stored or used for model training without the customer's consent. Terms differ between providers and between free and paid tiers, so read the actual contract for the tier you buy.
Then decide who checks the output. ISO 18587 covers full human post-editing of machine output and post-editor competences; it is a reasonable reference when a vendor offers "MT plus post-editing". For where raw MT is and is not acceptable, see machine vs human translation for Korean marketing.
Other tools you will hear about
- QA checkers: compare source and target for numbers, tags, terms and untranslated text. Often inside CAT tools.
- Terminology management: a central termbase shared across vendors and teams; can be stand-alone or part of a TMS.
- Quality estimation: scores MT output so that low-risk segments can skip or receive lighter human editing. Useful at volume; verify the routing on your own content before relying on it.
- Speech and device translation: real-time interpretation apps and earbuds, a separate category with its own limits.
Buyer's checklist
- Name the job: helping translators (CAT), moving content (TMS), or drafting automatically (MT).
- List every place your Korean content comes from and check each has a connector.
- Confirm you can export your memory (TMX), bilingual files (XLIFF) and terms (TBX or CSV).
- Read the data terms for the exact tier you will use, including MT built into other tools.
- Test MT with your own glossary on twenty real Korean segments, reviewed by a Korean linguist.
- Decide who reviews output, and budget their time, before the software goes live.
Short answers
Do we need our own CAT tool if we use an agency?
Usually not. The agency's translators use their own. What you need is ownership of the outputs: ask for your translation memory as TMX and your termbase in a standard format.
Is a TMS worth it for one language?
If Korean content changes weekly and comes from several systems (app, website, help center), often yes, because the automation saves more than it costs. If you send a few documents a quarter, email and an agency portal are enough.
Can we use free machine translation for confidential text?
Check the provider's terms first. Business tiers of some providers state that texts are not stored or used for training without consent; free consumer tools may have different terms. Your legal team should decide, not the person who happens to have the browser tab open.
Choosing tools for Korean?
We can test an MT engine and glossary on your content and tell you where human review is still needed. Contact us →
Related reading: Netflix localization lessons for brands