Concepts and terms
This page explains the terms used in the app and this guide. Follow the linked instructions when you need to perform an operation.
Project, file, and segment
A project is the job record. It brings together files, languages, client and job details, instructions, and applicable resources. A project can contain several files.
A file is one document or bilingual file within that job. Opening it loads the text into the editing workspace. A command that affects the current file does not necessarily process every other file in the project.
A segment is a piece of text used for translation or review. It is often a sentence, but it can be a heading, a label, or another part of a document. Segmentation rules and the source format determine the initial boundaries.
For example, a manual can be one file in a larger project, and its heading “Installation” can be one segment. Segments do not have to be complete sentences.
See Project Home, Translation table, and Segmentation.
Source, target, and translation unit
The source is the original text. The target is its translation. A translation unit pairs a source segment with its target; in a bilingual table, that pair occupies one row. Some controls use the shorter word unit.
In a single-language review, you revise the document's existing text in one working column. Do not expect every source–target tool to be available there. See Review projects.
The active segment is the row currently being worked on. The active side is the source or target editing area used by a command. A text cursor in the target and a selection of several table rows convey different information.
Selection, filtering, and scope
A text selection highlights wording within an editor. A row selection identifies one or more table units. A filter controls which rows are visible. Filtering does not delete the hidden rows or automatically select all results.
An operation's scope is the set of content it processes. Depending on the command, that can be selected rows, filtered rows, the loaded file or View, the project, or a phrase in the active editor.
For example, selecting two rows does not restrict Resegment to those rows: Resegment uses the current file. A QA profile also has a different scope from dictionary Spellcheck. Check the operation-by-operation scope guide before a batch action.
Clear filters to return to all loaded segments
Clear all on the translation filter bar removes the current table filters and related restrictions. It shows all segments in the open file or View. To see other project files, open them separately. To see segments outside a View, open their original file.
A project View
A View is a saved set of project segments, possibly drawn from several files. It lets you work on that set without creating separate copies of the translations. Changes affect the original units.
A View saves its membership when created. It is not a live filter that continually adds new matching rows or removes rows when their status changes. See Create a View.
The View ribbon is different: it contains display controls such as font size, segment layout, and whitespace visibility.
Saving, confirmation, and locks
Saving preserves changes. In a project-file editor, text edits are saved as you work. Confirmation records approval of a segment and, in a bilingual project, saves the translation to the attached memories. Confirmation can also propagate a target to later identical source occurrences in the open file or View.
An existing Confirmed status can come from an import or qualifying memory pretranslation as well as manual confirmation. Check imported and pretranslated segments before treating them as reviewed; run QA separately.
A lock prevents editing operations that require unlocked content. A locked unit can still be visible or used as a resource. Check a pair's wording before relying on it, even when it is locked.
Memory and termbase editors handle saving differently from project files. Saved entries normally open confirmed and locked. Unlock the entries you want to change, edit them, then Confirm to save the changes and lock them again. See Memory locks and saving and Term locks and saving.
Finishing a project and exporting a file are separate operations from confirming segments. Finishing checks the relevant project state; exporting creates a downloadable output. See Translation and review workflow.
Translation memory, matches, and concordance
A translation memory, or TM, stores source passages together with translations for reuse. A memory unit is one saved source–target pair. Attach the appropriate memories to a project so its normal memory tools use the intended resources.
A match is a saved pair offered because its source resembles the current segment. An exact match has a score of 100. A fuzzy match is similar rather than exact, so differences need attention. Scores of 101 or 102 include additional context information; they do not mean linguistic quality above 100%.
For example, a previous translation of “Close the front cover” may help with “Close the rear cover,” but inserting it without changing the component name would be an error.
Concordance searches saved memory pairs for a word or phrase. Use it to investigate previous usage when you do not need a match for the entire current sentence. See Translation memories and Concordance.
Pretranslation fills eligible targets from available memory translations and, when requested, AI drafts. Review the chosen scope and confirmation options before running it. See Pretranslation.
Glossary, termbase, and terminology matching
A glossary, also called a termbase, stores terms and their equivalents for a concept. It can include examples, matching settings, and forbidden wording. It serves a different purpose from a memory of complete translated passages.
For example, a glossary can record the required translation of “pressure relief valve.” A memory can store a sentence explaining when that valve opens.
Terminology matching recognizes glossary wording in a segment. Exact, case-sensitive, whole-word, or fuzzy settings affect recognition. A forbidden term records wording to avoid; check which language's entry carries that flag. See Glossaries and matching behavior.
File Index is another resource: it lists recurring source expressions in the current file. It is not the same as a maintained termbase or a search of previous translation memories. See File Index.
Tags and formatting
A tag represents formatting, protected content, or document structure carried with the text. A paired tag has an opening and closing marker; a standalone tag represents an element on its own.
Move tags with the content they belong to when the translation changes word order. Hiding tags changes their display, not the underlying file. A missing tag and a malformed tag are different problems: one is absent, while the other's structure is invalid. See Handling tags.
QA, spelling, and false positives
Quality assurance, or QA, checks defined conditions such as numbers, tags, punctuation, repeated wording, or required terminology. A profile chooses a collection of checks. A finding is a reported possible problem to investigate.
A false positive is a finding that is acceptable in context. Dismissing it does not correct the text. Use Ignore for an exception in the results, or Add to dictionary when a spelling should be accepted in future checks.
Spellcheck checks spelling or proposes language corrections through its selected mode. It is separate from general QA and from checking whether the translation conveys the correct meaning. See QA workflow and Spellcheck.
Comments, tracked changes, and proposals
A comment is a note about text. Tracked changes record edits made to text, which you can accept or reject. An AI proposal is suggested wording that you can inspect before applying it.
Applying a proposal, accepting a tracked change, and confirming a segment are separate decisions. Check what the command actually changes. See Comments and tracked changes and Change previews.
An AI rule is a reusable instruction included in applicable AI requests. A QA rule checks a condition. Saving one does not create the other. See AI rules and Custom QA rules.
Raw words, weighted words, and recorded activity
Raw words count the analyzed source text. Weighted words apply chosen percentages to repetition and match categories for a pricing calculation. For example, 1,000 raw words at 20% contribute 200 weighted words. See Quotation analysis and pricing.
Statistics measures also differ: Productivity uses qualifying confirmed source words, while the year-long Translation activity grid records target-word confirmation activity. Recorded project value, invoice totals, tracked time, and calculated hourly value each use their own definitions. See Dashboard and statistics.
Local work, synchronization, backup, and export
Local work is saved in the current browser installation or desktop app on that device. Cloud work uses your account's cloud workspace when your plan includes it. Signing in elsewhere does not copy all local work to that device.
Synchronization reconciles project copies through the supported sync action. A backup is a downloaded copy of local workspace data for recovery or transfer. A document export produces a translated or bilingual file for delivery or continued work. These operations are not interchangeable. See Storage and backups and Files and export.