Automating Localization: A Study with Transifex Native & Amazon Translate

Mike Giannakopoulos
November 10, 2020
6 min read

When considering a modern way of development and releasing digital content, we are talking about ease and speed both for brand new content and for updates of existing content. This stands true regardless of whether we are talking about a newly developed feature in an application or a new blog post (like this one). This is well summed up in the famous sci-fi phrase “Content must flow”!The most complex encounter of such a “flow” has to do with application development. In order to support fast development and quick releases, the development community is using a workflow called CI/CD, which stands for Continuous Integration/Continuous Delivery/Continuous Deployment. A CI/CD workflow allows multiple people to work and release in parallel within the same application. With the CI/CD principles in place, developers have achieved the astounding reality that we have all experienced over the last decade.

Of course, when adding content localization into the mix also adds complexity in the aforementioned workflow, which means certain friction points need some grooming and tweaking in order for everything to work. Localization works on top of files that are hard to manage over a CI/CD workflow. To achieve good translation quality, additional context is needed; plus, there are some possible issues that you will only be able to after everything is live on your application.

Certain steps are being taken in the localization field to address the above issues, but we are still on an exploratory stage. Translation quality and speed are constantly being improved by amazing machine translation (MT) advancements in the field and the issues of a CI/CD workflow over localization can now be partially tackled via integrations.

The most common localization workflow involves working in discrete content iterations, where all localized content is grouped under a single batch and released at once. In this type of workflow, all localization actions are organized around this single batch release. This means that all localization tasks regarding content review are placed before the release and any additional fixes or errors will need to be included in a new, upcoming release. This process introduced is mostly linear, with certain steps needed to be made in a specific sequence. 

The good news is that, aside from this common localization workflow, there are other more and more technologies being built to help to overcome the many workflow challenges of localization. In this post, we will break down a specific new advancement: how Amazon and Transifex are enabling localization teams to handle the problems of shared workflows between the two services.

Amazon has entered strong in the MT field offering deep learning techniques to produce more accurate and higher quality translations. Moreover, Amazon Translate service supports a named entity customization where you can feed your custom terms and their industry specific translations into Amazon Translate and the translation will respect those terms and their translations. Those named entities are the true power that Amazon brings in a more automated localization, ensuring a big benefit over competing commercial MT services without the additional cost of a totally custom MT engine.

Transifex is working hard to build smart integrations that will enable companies to overcome issues around localization workflow automations, especially for application content. With a recent innovation, Transifex is trying to overcome all the issues introduced by localization files by launching a new solution called Transifex Native, that renders localization files obsolete by embedding localization directly in the code. With Transifex Native, localization becomes truly continuous and all content is managed within a single place, the Transifex platform.

Leveraging both Amazon Translate and Transifex, you can get a truly fast and high quality localization flow that moves as fast as your development team does. Both services are fully available using the Amazon Translate integration that connects your Transifex glossary terms with the Amazon Translate service. Setting up and using Transifex Native in your application code enables developers to mark and send content for translation to Amazon Translate and have glossary-aware translation available from Amazon Translate even faster than new features are released.

Setting up a localization workflow with Amazon Translate and Transifex, you can combine the benefits of higher quality machine translations using your industry specific glossary terms and custom translations along with an automated way to get content for translations and serve translated content in your application. A CI/CD workflow would look like this:

  1. Developers use Transifex Native SDK in code to mark content that needs translation.
  2. Developers sending content to Transifex once they consider it ready.
  3. Amazon Translate with Entity terminology informed by a glossary kicks in to translate content into selected target languages.
  4. Translated content is fed back to the code using Transifex Native.

The above workflow requires minimum custom intervention after being setup once (“set it and forget it”) with the benefits of full automation and high quality localization outcome that you and your team controls via properly training the Amazon Translate service. Also taking advantage of the real-time nature of Transifex Native, you can easily fix any translation errors identified in your application without the need for an additional release. Overall, localization moves as fast as development offering all advantages of a CI/CD workflow, where you get fast and high quality results with the additional control over improvements and updates.

Comparing the workflow when using Transifex Native and Amazon Translate with the common workflow presented at the beginning of this blog post, we can instantly spot some benefits:

  • Localization work is a continuous stream versus organizing tasks around a release that you can hit or miss,
  • Localization is alive and improved through fast iterations that are molding along with new technology over Amazon Translate and organizational improvements over new glossary terms and manual fixes on need basis.

The above workflow is available right now using both services. Amazon Translate is already fully integrated with Transifex leveraging all the benefits that AWS offers. An improvement over the available Amazon integration will auto-sync glossary terms between Transifex and Amazon automatically, further reducing the need for a human hand. Transifex Native is already available for Django and JavaScript, including react.js and node.js with more language frameworks to be supported in the near future.

— 

Interested in giving localization a try for your fintech company? Sign up for your risk-free, 15-day free trial of Transifex to see exactly how localization can help you level up your global fintech game: www.transifex.com/signup/.

Mike Giannakopoulos
FacebookgithubGoogle+Fill 88Twitter