Locale File Translation (Strings)

Locale file is one of the three file modes supported by Project Translation, built for key-value resource files (key = value). Transbox parses the file into segments keyed by their path, lets you translate them one by one, and writes them back into the original structure on export.

Supported Formats

The same extension can mean different things per mode: .xml is parsed as Android strings in Locale file mode but as generic XML in Document mode; .xlsx is parsed as Key/source columns in Locale file mode. The mode you pick when adding a file decides which parser runs.

Parsing Rules

• Key-based extraction: each leaf string becomes a segment and keeps its path (e.g. navigation.home), which is restored on export
• Type preservation: JSON number / boolean / null values keep their type and are written back as values, not strings
• Placeholders and inline tags: {name}, %s, %1$@, <b>, <a href> are treated as protected chunks and rendered as tags in the editor so they are not accidentally deleted
• Duplicate detection: segments with identical source text are flagged, and can be filtered as "first occurrence", "other duplicates" or "not duplicated"

Workflow

1. Create a project

1. Smart Translation → Project Translation
2. New project: name, notes, source language, target languages
3. You land on the project's file list

2. Add locale files

1. Click Add file inside the project
2. Pick the Locale file mode
3. Choose the file, confirm source and target languages
4. The file list shows the key count and update time

3. Translate in the editor

Click the file name or Translate to open the editor, which has four layers:

• Header: file name, source/target language switcher, version log, preferences, submit, export, import translations
• Toolbar: search (key/source/target), stage filter, more filters, batch operations, QA, file analysis, pre-translate, re-parse
• Three columns
  • Left: segment list (index, source, target, status marks)
  • Middle: the target editor for the current segment, with Edit history / QA / Comments tabs below
  • Right: vertical tabs for Corpus (TM → termbase → MT), Add term and AI assistant
• Status bar: progress, paging and other state

4. Export

Export in the header (or More → Export in the file list) produces:

• Translated file: original format, original key structure
• Excel comparison: source and target in two columns
• XLIFF: for other CAT tools or external translators
• JSON: a uniform comparison structure

The scope can be narrowed by file, language and key conditions (search text, stage, marked, duplicate type, change source).

5. Import translations back

Import translations accepts .json / .xlsx / .xliff / .xlf with options:

• Auto confirm: mark imported segments as confirmed
• Write to memory: store the imported pairs as memory entries
• Push to stage: place segments at translation / review / proofreading / published

The project-level Import translations handles several files at once.

Translation Aids

Memory and termbase matching

The right-hand Corpus panel shows memory matches, termbase matches and MT results in order, each applicable in one click. Matching refreshes whenever you switch segments.

Pre-translation

Pre-translate processes the untranslated segments in the current filter:

• TM first: use memory matches, fall back to MT for misses
• TM only: memory only, no translation cost
• MT only: machine translation for everything

Set the minimum TM match rate (default 70%). A progress bar runs and can be stopped at any time; the summary reports TM hits, MT results and failures.

Batch operations

Batch operations apply only to the current filter:

• Update stage (target stage + confirm flag)
• Clear translations
• Copy source to target
• Copy the translation to another language
• Run QA on the range
• Smart tag fill (complete inline tags in target following the source order)

Each run asks for confirmation and shows progress.

QA checks

QA compares structural elements between source and target: missing/extra placeholders, HTML tag count and pairing, missing/extra variables, leading/trailing whitespace, punctuation, number counts, special characters, terminology consistency, URL count and modification, newline counts, and inline tag presence/order/pairing/nesting.

Results are listed in the QA tab of the middle column and can be ignored or restored individually; unignored issues count toward the segment and project QA totals.

File analysis

File analysis reports, for the current target language: segment count, source/target character counts, translated / untranslated / confirmed / MT / QA segment counts, internal repetition (total, propagable, group count), applied TM match distribution (100% / 90-99% / 80-89% / … / unmatched) and stage distribution — useful for judging delivery status and workload.

AI assistant

The AI assistant tab opens a session scoped to the current segment and target language: explain the source, propose alternatives, check tone. The project-level AI prompt in Project settings acts as the system prompt for every segment in that project. More → Agent translate in the file list lets the agent translate a whole file following the project prompt.

Shortcuts

Notes

1. Never drop protected tags: placeholders and inline tags in the target must match the source, or both the application and QA will complain
2. Re-parse after source changes: when keys are added or removed, use More → Re-parse; existing translations are kept by key
3. Plan limits: files per project and characters per file are capped
4. Filter before export: to deliver only reviewed segments, filter by stage first
5. Submit advances the review round: Submit increments the file's pending review version, marking the current segments as submitted; use the version log to review the change history of the file and its segments

FAQ

Q: Why is the "Locale file" card greyed out?

A: Your plan does not include locale files. Upgrade the plan, or add the file with the Document mode.

Q: The key count is lower than the number of entries in the file. Why?

A: Only leaf strings are counted. Nested objects are containers, and comments or empty objects are not segments.

Q: Some exported values turned into strings.

A: Check whether the original value was a number/boolean/null. If the source was a quoted string, the export keeps it a string.

Q: Does re-parsing lose translations?

A: No. Re-parsing aligns by key and keeps existing translations; keys that disappeared are dropped.

Q: Can it extract strings directly from source code?

A: Not currently. Generate a resource file with your existing i18n extraction step, then add it to a project.

Q: How are multiple target languages handled?

A: One file can carry several target languages; switch languages in the editor header. Use Copy translation to another language to seed a draft in the other languages.