Translation Reasoning
Once a segment has a Context Sensitive Translation, a Translation reasoning switch on that card shows the rationale behind the AI's translation as a short set of categorized bullets — why a specific glossary term was used, why the wording was adjusted for fluency, and so on.
💡 Who is this for? This guide is for any user with Editor access who needs to turn on and read Translation reasoning to verify or understand a Context Sensitive Translate suggestion within the wxrks platform.
What Translation reasoning shows
Turning on Translation reasoning reveals why the AI made the choices it did for one specific Context Sensitive Translation — not a general explanation, but a breakdown tied to this segment's actual wording. Each reasoning bullet names a category and gives one concise sentence citing a concrete word or phrase from the segment, so you can trace a specific translation choice back to a specific cause instead of guessing.
In the example above, Context Sensitive Translate used Traduza — the exact inflected form stored in the account's glossary for "translate" — and the reasoning bullet spells out exactly that: a Glossary Enforcement decision citing the glossary's own target term. Only the factors that actually influenced that particular translation are shown; a segment with no glossary hit and no notable fluency adjustment might show a different category, or none at all.
Turning Translation reasoning on
Open a segment that already has a Context Sensitive Translation (or trigger one — see Context Sensitive Translate).
In the Suggestions tab, find the Translation reasoning switch directly on the Context Sensitive Translation card.
Turn it on. The categorized bullets appear underneath the switch.
Translation reasoning is your own Editor preference, separate from the Enable Context Sensitive Translate and Enable Review Assessment switches in the Editor's Settings panel, which control whether Context Sensitive Translate runs at all. It has no separate permission or admin setting behind it — it's off by default for every user, and once you turn it on, it stays on for you across segments and sessions until you turn it off again. Turning it on for a segment that already has a translation refreshes that translation immediately so the reasoning appears; turning it on with no existing translation just saves your preference and takes effect the next time Context Sensitive Translate runs.
What each reasoning category means
wxrks's AI is asked to prefer one of nine standard categories for each reasoning bullet, so the labels stay consistent across segments, languages, and projects instead of drifting into ad-hoc wording. Only categories that actually apply to a given segment are shown.
Category | What it means |
Glossary Enforcement | A specific word or phrase was chosen because it matches an approved glossary entry. This is the only category that also cites the exact glossary source and target term used, so you can check the substitution against the glossary yourself. |
TM Reuse | Wording was carried over from a matching or closely-matching Translation Memory segment instead of being translated from scratch. |
Fluency Adaptation | Phrasing, word order, or grammar was adjusted so the sentence reads naturally in the target language, even where that means departing from a literal, word-for-word rendering. |
Audience/Tone | Word choice or register was shaped by the audience, tone, or formality instructions set for the segment (e.g. more formal phrasing for a corporate audience). |
Terminology Choice | A specific term was chosen among several plausible options for domain or subject-matter reasons — distinct from Glossary Enforcement, which is reserved for an actual stored glossary hit. |
Style Adjustment | Sentence structure or voice was adjusted to match a style convention for the content, separate from a pure fluency fix. |
Localization | Content was adapted for the target locale's own conventions — units, date/number formats, or a culturally-specific reference — beyond a direct word choice. |
Error Correction | A mistake or inconsistency — in the source text or in an earlier draft translation — was identified and fixed. |
Formatting | Tag placement, punctuation, or spacing was adjusted to fit the target segment's formatting rules. |
A reasoning bullet can occasionally use a different, non-standard label if none of the nine above genuinely fits — the AI is asked to prefer this list, not restricted to it, and always uses the same label consistently for the same kind of decision.
Why turn this on
Translation reasoning is entirely optional — Context Sensitive Translate works the same with it on or off. It's worth turning on in a few specific situations:
Spot-checking AI decisions before approving a segment. Instead of re-deriving why the AI chose a particular wording, the reasoning tells you directly — so you can confirm it's sound rather than guess.
Understanding a surprising glossary substitution. If a translation uses an inflected or unfamiliar-looking form of a glossary term, a Glossary Enforcement bullet shows exactly which glossary entry it came from, so you're not left wondering whether it's a real term or a mistake.
Training new team members on what "good" AI output reasoning looks like. Reviewing a few segments with reasoning on is a fast way to show a new linguist or reviewer how the AI is meant to be using TM, glossary, and fluency signals together, and what a well-justified suggestion looks like versus one that needs a closer look.
A segment can surface more than one category at once, as above — here both Glossary Enforcement and Fluency Adaptation influenced the same translation, and both are listed because both genuinely applied.
Translation reasoning and Augmented Term Lookup
Translation reasoning explains this Context Sensitive Translate suggestion — it's a separate feature from Augmented Term Lookup's own glossary-variant suggestions, which run independently of Context Sensitive Translate and are never applied to a translation automatically. Augmented Term Lookup's GLOSSARY VARIANTS panel runs on its own as soon as you open a matching segment, whether or not Context Sensitive Translate has run; and Context Sensitive Translate's own output is never changed by what Augmented Term Lookup suggests. In practice this means a segment can show two different forms of the same word at once — Context Sensitive Translate's own choice, explained by Translation reasoning if it's on, sitting alongside a separately-offered, unapplied Augmented Term Lookup variant.
Troubleshooting
The Translation reasoning switch doesn't appear. The switch only renders once the segment already has a Context Sensitive Translation — open or trigger one first (see Context Sensitive Translate). If Context Sensitive Translate itself isn't available, check with your Account Admin that Enable Augmented Translation Actions is on for your account and Organizational Unit.
No bullets show even with the switch on. A blank result means the AI genuinely found no factor from the standard list that influenced that particular translation — try Refresh on the Context Sensitive Translation card to have the AI re-evaluate the segment with reasoning included.
Still stuck? Search prior questions or ask the wider user base on the wxrks Community forum, or reach out to [email protected].
Quick reference
Question | Answer |
Where is the switch? | On the Context Sensitive Translation card itself, in the Suggestions tab — not in the Editor Settings panel. |
Who can use it? | Any user with Editor access — it's a personal preference with no admin setting or extra permission behind it. |
Default state? | Off. Once turned on, it stays on for you across segments and sessions. |
Does turning it on change the translation? | Only if the segment already had a translation — turning the switch on refreshes it immediately so the reasoning appears. |
Related articles
Context Sensitive Translate — the AI translation action Translation reasoning explains.
wxrks Augmented Actions — the full hub article for Context Sensitive Translate, Alternative Suggestion, Proofread, Check Smells, Fix Tags, and account/Organizational Unit-level setup.
Augmented Term Lookup — a related but independent AI feature that suggests inflected glossary term variants; see "Translation reasoning and Augmented Term Lookup" above for how the two relate.