We are happy to announce the beta release v0.9.1 of the Moses for Localization open source project! You can download the packages on the project website: http://code.google.com/p/m4loc/downloads/list
The Moses for Localization (M4Loc) project grew out of the desire to close gaps for the use of the Moses open source MT engine in a language industry setting with open source. M4Loc provides tools to process localization-specific formats in Moses and to integrate Moses in localization workflows. (for more information on the background see this article <http://www.translationautomation.com/best-practices/want-to-ride-the-machin e-translation-tidal-wave.html> and this presentation <http://www.digitalsilkroad.net/m4loc.pdf> ) What is New With this Beta release we are adding support to translate XLIFF (XML Localization Interchange File Format) with Moses and now have all the tools in place to address the major gaps. Adding XLIFF support would not have been possible without major support from the Okapi project (http://okapi.opentag.com/). Yves Savourel added a filter to the M10 release of Okapi that allows translating XLIFF and other file types with the M4Loc/Moses tool chain. For a complete list of file types see: http://www.opentag.com/okapi/wiki/index.php?title=Filters For a complete list of M4Loc features see this wiki page: http://code.google.com/p/m4loc/wiki/Features Getting Started You can get started by downloading the package appropriate for your platform (http://code.google.com/p/m4loc/downloads/list) and following the instructions in xliff/INSTALL.txt and xliff/README.txt. There is also a detailed tools reference available on the wiki: http://code.google.com/p/m4loc/wiki/ToolsRef We included a pseudo-translation script that allows you to go through the whole process without having a trained Moses system available. Here are the steps translating one of the test XLIFF files on Windows (these steps assume that Okapi and M4Loc are correctly set up): 1. Open a command window and change into the xliff directory 2. xliff2moses.bat .\t\languagetool.xlf en-us This extracts the source segments from the XLIFF file and places them into a plain text file languagetool.xlf.lcs.en-us which is translateable with Moses or the pseudo-translator. 3. perl pseudo_translate.pl -n 2 < .\t\languagetool.xlf.lcs.en-us > .\t\languagetool.xlf.ucs.fr-fr This pseudo-translates the English source text into Pig Latin (with a maximum phrase length of 2) and emits phrase alignment information between the source and target sentence. The translations are not correctly uppercased here, but we pretend that they are for the purpose this example (lcs stands for "lower case" and ucs for "upper case"). 4. moses2xliff.bat .\t\languagetool.xlf en-us fr-fr This command merges the translations back into XLIFF as <alt-trans> elements. The file .\t\languagetool.out.xlf can now be imported into a translation environment for post-editing. What we are looking for For now we designed M4Loc intentionally as a tool box, a collection of scripts. This allows for maximum flexibility when integrating these tools into existing localization workflows - ranging from commercial or open source TMS systems, online APIs to content management systems for online gisting translations. We would like to hear from you if M4Loc addresses your integration needs. XLIFF is a very flexible and extensible standard, which is why different tools output a wide variety of XLIFF dialects. Does the M4Loc/Okapi combination work with the XLIFF dialect your tools are using? M4Loc maintains inline formatting for translated segments - is the quality of this formatting sufficient for your purpose (e.g. gisting, post-editing)? We believe the quality compares favorably to other approaches - should there be an objective measure supplementing traditional MT evaluation metrics? Please send feedback to the project maintainers (in the To: line) or even better to the M4Loc mailing list: http://groups.google.com/group/m4loc - if you are not a mailing list member yet, we invite you to join. Support Community support is available through the mailing list: http://groups.google.com/group/m4loc Digital Silk Road offers commercial support. Acknowledgments I would like to thank Yves Savourel for the additions to Okapi making M4Loc possible and his outstanding responsiveness to our requests. Also many thanks to Tomáš Hudik for the many hours spent on coding, testing and documenting for the project. Many thanks to the project sponsors Moravia, TAUS and Digital Silk Road! The project is open to additional contributors and sponsors. If you find the project useful, please consider donating with the PayPal button on the project front page.
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
