Skip to main content

Upload a file

Drop a bilingual file on Run a check, or on the drop area Home shows until the workspace has its first report. Check your first file walks through the first run.

No file to hand? Try the sample. The rule checks run first, then the AI reads every translated segment. You get one report with every finding, a score and a result; nothing in your file is changed. Every run keeps the profile version and the rule set that produced it.

Home with five reports in 30 days, three of them passing: the clean-rate line at 60%, the newest reports with a line of what needs you under them, where the findings come from and the profiles with the most criticals.Home with five reports in 30 days, three of them passing: the clean-rate line at 60%, the newest reports with a line of what needs you under them, where the findings come from and the profiles with the most criticals.
Home once a workspace has five reports in 30 days: the clean rate, the newest reports, what needs you and where the findings come from. Northwind and Contoso are demo accounts.

Accepted files​

XLIFF in any of its flavours (.xliff, .xlf, .sdlxliff, .mxliff, .mqxliff, .txlf, .xlz, .wsxz), a Trados or memoQ package, and a bilingual spreadsheet (.xlsx, with the columns confirmed). Software string files: gettext .po on its own; i18n .json, .NET .resx, iOS and macOS .strings and .stringsdict, Java .properties as the source file and the translated file together, or in a zip. The report names each string by its key. The format is read from the file's content, not its extension. Up to 200 MB per upload.

Software string files​

A gettext .po file holds the source (msgid) and the translation (msgstr), so it is checked on its own. Plural forms are checked one by one, a fuzzy entry is marked for review and obsolete entries are skipped.

An i18n .json, .resx, .strings, .stringsdict or .properties file holds one language. Drop the source file and the translated file together, for example en.json and tr.json, or a zip with both. The files are paired by the language in their name or folder. Each line below is one pair:

  • en.json and tr.json
  • common.en.json and common.tr.json
  • strings_en.json and strings_tr.json
  • locales/en/common.json and locales/tr/common.json
  • Resources.resx and Resources.tr.resx
  • en.lproj/Localizable.strings and tr.lproj/Localizable.strings
  • messages.properties and messages_tr.properties

The source is the file in the profile's source language, else the file with no language in its name, else the English file. The report says which file it checked against which.

A key in the source file with no translation is reported as untranslated. Keys only in the translated file are counted in one warning and are not checked. A single one-language file on its own is refused with "This is a single-language file. Drop the source and the translated file together, or a zip with both."

What each format brings:

  • .resx: each data entry is one string, named by its name, and its comment is the note. Files, images and typed values (a size, a colour) are skipped and counted.
  • .strings: each "key" = "value"; line is one string and the comment above it is the note. UTF-8 and UTF-16 files are both read.
  • .stringsdict: each plural form is one string, named by its key and #plural= with the form. The forms of the translated file's language are checked against the source's forms, so a Polish few is checked against the English other.
  • .properties: key=value, key: value and key value lines, lines continued with a backslash and \u escapes are read the way Java reads them. The file is read as ISO-8859-1, the Java rule, unless it is valid UTF-8 with characters beyond ASCII.

Placeholders are read the way the software reads them, ICU message arguments included: %s, %@, %1$@, %d, {0}, {name}, {{name}}, ${name} and %(name)s. In {count, plural, one {# file} other {# files}} the placeholder is {count}; the plural keywords and the text of each branch are not placeholders. A plural form for one number may leave the count out.

Several files​

Drop several at once. Files with the same profile are checked together as one report, which shows in Reports as the first name and "+N more". Files with different profiles run as separate reports, one per profile, each with its own profile, the way the Dropbox watcher handles each file.

The settings for one check​

Under the file's facts: the profile (Auto-detect, or pick one), AI review, the AI check of rule findings, the matches and locked segments to skip, spelling and LQA. They override the workspace and profile defaults for this run only.

Skip has three separate boxes. 100% skips exact matches only. ICE skips in-context exact matches, also called 101% or context matches. Locked skips locked segments. A skipped segment is not checked at all and the report says how many.

The same file again​

Run a new version of a file you checked before and the report starts with last time's decisions carried over and says what was resolved, what remains and what is new. A re-run of the same file within 30 days is free.

Try a sample file​

Try a sample file on the Run page runs a demo file so you can see a report before you have a file of your own.