Difference between revisions of "Projects/Summer of Code/2007/Projects/KAider"

Line 61: Line 61:
* [http://www.triplespin.com/en/products/locfactoryeditor.html#whatsnew locfactoryeditor] --Mac, $30
* [http://www.triplespin.com/en/products/locfactoryeditor.html#whatsnew locfactoryeditor] --Mac only
* [http://www.heartsome.net/EN/downloads.html Heartsome] - multiplatform, costs money
==KBabel features to be implemented==
==KBabel features to be implemented==

Revision as of 23:01, 26 November 2007

KAider is a computer-aided translation system that focuses on productivity and performance. Translator does only creative work (of delivering message in his/her mother language in laconic and easy to understand form). KAider implies parapgraph-by-paragrah translation approach (when translating documentation) and message-by-message approach (when translating GUI). See KAider/Introduction

Current state

Already has:

  • syntax highlighting
  • spellcheck (sonnet needs improvement)
  • search-n-replace, ignoring accel marks
  • formats .po file output better so less diff is generated by scripty
  • small features like quick tag insert, placing text cursor right after the tag in the beginning (e.g. '<qt>|foobar</qt>' where "|" is a cursor), entry bookmarks
  • viewer of the difference between current msgid and previous one (i.e. msgid translation of which current msgstr really is -- for fuzzies generated with --previous gettext option)
  • merge mode for editors (QA) or when several translators work on the same file screenshot
  • basic projectmanager functionality screenshot
  • Translation Memory (threaded) with shortcuts for inserting suggestions into current 'msgstr', scores are computed based on common/total length ratio, removed+added length, and count of removed+added parts screenshot
  • for difference representation in all places word-by-word algorithm is used (based on the Longest Common Sequence o(n*n) algorithm and my own experience)
  • glossary with basic tbx format support. KAider displays relevant entries on-the-fly and provides shortcuts to insert them. also, you can add new glossary terms via context menu of the glossary screenshot
  • webquery view, flexible thanks to kross
  • Search/Replace functions in multiple files at once.
  • Spellchecking of multiple files at once.


After you set kde env up (compiling kdelibs is enough):

cd trunk/extragear
svn up sdk
su kde-devel
mkdir sdk/kaider/build
cd sdk/kaider/build
cmakekde ..

as a root, run sshd and then from the usual shell:

ssh -XC [email protected]

you can get catalogmanager by specifying --project option

kaider --project /path/to/index.ktp

See http://websvn.kde.org/trunk/l10n-kde4/ru/


  • Majority of actions must be accessible via keyboard (because it is faster to press a shotcut than to be frustrated with a mouse)
  • Do automatization _everywhere_ it possible
  • Focus on quality. This is open source -- so source code is available (for change)

Further work

  • dbus
  • WebQuery for twin languages (like ukrainian and russian)
  • xliff+qt-linguist support (see KBabel features to be implemented)
  • Automatic Glossary building
  • Research on dividing into sentences rules (e.g. srx)
  • project-wise and program-wise: webquery scripts, glossaries, TMs
  • Glossary checklists: check for forbidden terms in new translation
  • Tighten SVN support: svn diff-like feature
  • Automate submitting translation suggestions to translate.google.com [Kross action]
  • back-checking, to see whether a term/string in the target language has been used as translation for different things, not just that one source has the same translation everywhere.
  • [Kross (project) action] that merges translations with updated template
  • fill TM with content of /usr/share/locale/<lang>/LC_MESSAGES/*.mo [Kross action]

Not for KDE:

  • Be complete computer-aided translation system by providing e.g. actions to import+export openoffice, txt and documents of other formats by calling appropriate scripts/commands. Define for that general kross actions interface.
  • Make nice windoze package for the windowzerz


KBabel features to be implemented

...in a smarter way :). After or during the summer.

  • persistent bookmarks for messages in a file saved in the project
  • extended marking of .po and .pot files (e.g. translator that currently works on the file and cince when) saved in the project
  • Opening source code by references in message comments [Kross action]
  • A plugin framework for validation tools for consistency checks [Kross action triggered on saving]
  • Sending the file using email [Kross (project) action]
  • Automatic syntax check with msgfmt when saving and, if an error occured, easy navigation to messages, which contain errors. == Syntax check (msgfmt --statistics) for existing files to control if the translated files will compile and, accordingly, work when distributed [Kross (project) action]
  • CVS and SVN support [Kross project action] (is 'svn ci' so hard?)
  • statistics
  • PO File Header change [Kross action (+triggered on saving)]
  • Printing of selected messages (eg fuzzy ) [Kross action]


  • msgid-diff-msgstr from [1] (features for all other commands are already implemented, if you haven't noticed this)

KBabel features NOT to be implemented

  • Character selection tool integration, sort by the frequency

Why? Better improve system-wide charselect tool, OR... modify your xorg keyboard layout!

  • Automatic ("rough" in kbabel terms) translation.

Why? because sometimes one English string may have two or more different translations depending on the context.

What I'm going to do is implement _interactive_ (or message-by-message) rough translation. If the message is already translated somewhere else, it suggests the translations (several, not one!) and displays them in the helper window. Translator may then choose one of the translation suggestions by pressing ctrl+1, ctrl+2, .. or ctrl+9, which will immediately insert it into msgstr  (replacing the old translation if it exists).

What old rough translation didn't provide is the ability to choose.

UPDATE: ok, i'll probably allow translator to autopopulate catalog with:

  • one-choice suggestions (e.g. if there are only one 100% suggestion), optionally mark fuzzy
  • the most recent suggestions for the cases of several 100% items, mark fuzzy


  • Create project, saving *.ktp file to l10n-kde4/<LangCode>
  • Populate Glossary via GlossaryView context menu (.tbx file will be created automatically for you on the first entry addition).
  • Populate Translation Memory by dropping .po files onto TM View

Content is available under Creative Commons License SA 4.0 unless otherwise noted.