Distributed Translation - Building Multilingual Publishing Systems

From Ott09 Wiki
Jump to: navigation, search

Natural evolution of content HURTS multilingualism

  • It's important to think about how a publication evolves and grows.
  • Commonly it starts out as focussed thing, particular topic or scope.
  • At first you don't think about how to make it multilingual and avilable to whole world because you have a specific audience.
  • Growing organically means starting with simple tools which traps you later when you want to expand to more languages.
  • e.g. of Twitter, can't scale linguistically because they didn't think of it at the beginning.

Most Basic Solution: Duplicate site in multiple languages

  • Simplest thing to do is set up multiple websites and just label them as translations, paste in content translations
  • e.g. GV or TechCrunch or Wikipedia
  • Easy to set up at first, doesn't require specialized CMS.
  • Slow and steady stream of content is available, fully segregated by language.
  • No rush to publish or to publish everything in all languages.

Integrated Multi-lingual CMS

  • note: post content multilingualism is seperate from template or UI translation.
  • Joomla and Drupal allow you to have child documents that are treated as versions of parent original in another language.
  • TikiWiki does this with content staging and validation. * Works but lacks built-in workflow, needs a lot of editorial oversight to keep clean and functional.
  • Workable if you have a well organized site and not too much content and strong editorial.

External Translation management platform

  • create a DB of documents that can be translated and keep their translation status organized seperate from the CMS
  • E.g. monitor rss feeds, pull in data for central queue.
  • Good if you have a lot of content and want to keep the translations managed.
  • As a site grows you end up with several sites maybe with multiple layouts. So you end up with specialized sites that have their own needs. Having the translations managed externally allows freedom of running each site in its own environment.
  • integrating the translation manager with your CMS is an extra level of work that may/may not be easier than setting up the translation inside the CMS.

Embedded translation memory with fallback to machine

  • Translation Memory embedded in content to multilingualize automatically but let humans fix it
  • Doesnt' need to know about the content or CMS
  • harder to do, modifications to website are necessary
  • simple and flexible to deal with translations.
  • I.e. WWL: hybrid system of machine and human translation
  • allow users to do translation but with filters and permissions to control.
  • good for making news avaialable immediately with machine translation and letting human translations come later.
  • set up fallbacks so better version is shown if available otherwise machine.
  • Good for high volume site with a lot of posts, time sensitive material.