Words we use
Our industry has a lot of shorthand. Here is all of it that appears in this product, in plain words.
Files and text
| Word | What it means |
|---|---|
| Segment | One unit of text, usually a sentence. Translation tools cut a document into segments and translate them one at a time. Findings are reported per segment. |
| Source and target | Source is the original text. Target is the translation. Every check compares the two. |
| Bilingual file | A file that holds both, side by side. That is what this tool needs. A finished Word document is not bilingual, because the original text is gone. |
| XLIFF | The standard file format for bilingual files. Most translation tools can export one. The variants (.sdlxliff from Trados, .mxliff from Phrase, .mqxliff from memoQ and others) are all read the same way. |
| Software string file | The file an app keeps its text in, one string per key. A gettext .po file holds the source and the translation. An i18n .json, .NET .resx, iOS and macOS .strings or .stringsdict or Java .properties file holds one language, so it is checked as a pair: the source file and the translated file. The report names each string by its key. |
| CAT tool | Computer-Aided Translation tool: the editor a translator works in. Trados, memoQ, Phrase, XTM and Smartling are CAT tools. |
| Tag, placeholder | Code inside the text that is not text: bold markers, links or a slot like {name} that the software fills in later. If one is lost or changed in translation, the page can break or show the wrong value. This is the most common critical finding. |
Reuse and matches
| Word | What it means |
|---|---|
| TM (translation memory) | A store of everything translated before. Tools reuse it, so a lot of a new file may already be translated. |
| 100% match | A segment the memory already had, word for word. Usually somebody else's earlier work, so usually not the current reviewer's responsibility. |
| ICE, context match | A 100% match whose surrounding text also matched, so it is even safer. ICE stands for In-Context Exact. |
| Locked segment | A segment the project manager has closed for editing. You cannot change it, so a finding on it is usually not yours to fix. |
| Leverage | How much of a file came from memory rather than being newly translated. |
This is why the run options let you exclude 100%, ICE and locked segments: it leaves the segments the reviewer is answerable for.
Rules and terminology
| Word | What it means |
|---|---|
| Profile | Everything the tool knows about one client and one language pair: which checks matter, which words must stay in English, which must be translated a set way and how serious each problem is. |
| DNT (do not translate), Brands and terms to keep | Words that must stay exactly as they are. Brand names, product names, UI labels. |
| Glossary | The opposite: words that must be translated one agreed way, every time. |
| Fine as is | Words that may stay identical to the source without being flagged as untranslated. |
| Severity | How serious a finding is: critical (must fix, fails the check), major (look at it, it counts toward the score), minor (for your awareness, no weight unless the workspace gives it one). You can change a check's severity per client. |
| False positive (FP) | A finding that is not really a problem. Marking one with Ignore records it on that report only. To stop it coming back, teach the profile through Learn or change its rules. |
Scored review
| Word | What it means |
|---|---|
| LQA (Linguistic Quality Assurance) | A human review that produces a score, not just a list: a reviewer classifies each error by type and severity, and the scorecard turns that into a number you can track or pass or fail on. |
| MQM, DQF | Published standards for those error categories, so a score means the same thing to two different companies. A scorecard template can follow one of them. |
| Scorecard template | The categories, weights and pass mark a review is scored against. Set once, reused for every review of that type. |
How this tool talks about itself
| Word | What it means |
|---|---|
| Check, run | One file going through the tool once. Every run is kept, with its report. |
| Finding | One thing the tool noticed, on one segment, with the evidence for it. |
| Rule engine | The part that checks things with certainty: tags, numbers, terminology, spacing. It does not guess, and it is instant. |
| AI review, AI pass | The second layer. It reads meaning, so it can catch an omission or a mistranslation that no rule can describe. It also checks the rule engine's findings and marks the ones it disputes. It never changes your original file. |
| AI check of rule findings | The AI's verdict on a rule finding: AI agrees or AI disputes, with its reason. A disputed finding stays in the report and counts at half weight until you decide. |
| Second opinion | A second model from another provider rechecks the AI review's critical and major findings and confirms the ones it agrees with. Off by default. |
| Confirmed | A finding the AI confirmed or an exact rule check found. |
| Quality score | A number out of 100 the tool calculates from the findings and their severities. It sits beside the result, not inside it: a file can read FAILED with 2 critical and Quality 95.5/100 at once, which means a good file with two things that must be fixed. |
| Learn | The queue where reviewer feedback becomes suggestions on a profile, approved one by one by a person. |