Using AI for innovative language translation
When people come calling to request if they can translate your work into their heart language, signs are good you’re on to something with the content you’re creating. This has been the story of The Word One to One over the last year or so. Having published a series of books guiding readers through the book of John in the Bible many years ago, and then developing that into a first-class digital product too, demand has grown across the world for translations.
Translation work is hard, and with a sensitive subject matter that requires careful translation relating to meaning and interpretation, the effort is further compounded, and each request feels like another mountain to climb.
Yet, with requests inbound, it’s important to capitalise on the engagement and enthusiasm of a wide user base, and the opportunity to reach more people in more language communities.
Challenge
How might we enable rapid, safe and scalable translation to allow Word One to One to continue it’s organic growth and at a rate the world demands? Having worked closely with the Word One to One team for a couple of years on the digital transformation of their work and bringing it to life through a unique interactive app experience, the attention quickly turned to the translation endeavour.
The Solution
Building on scout+redeem’s solid publishing platform, we set to work understanding the challenges, nuances and process such a translation would take. Content falls into several main parts, the Bible itself, where a suitable existing translation should be identified, the core Word One to One notes themselves, where careful understanding of context and objective are key, and finally the app copy that guides the user through the experience.
Through our Design Sprint process, engaging team members and contributors from several countries from sub-Sahara Africa, to Central America and with the UK we designed a process to translate, moderate, test and validate translations.
Working in an agile manner we developed the first set of tools to enable seamless side by side translation across the complex content structure, with both plain text and real time in app previews available at every key stroke.
Since we developed the app experience in Flutter, we were able to bring the full app experience directly inside the editor for instant visual feedback too. Combining Flutter web with our Node JS web platform might just be the first time a translators has been able to see their translation drop into live context in real time.
The next step was to find ways to accelerate the process without compromising quality. Enter Lilt AI platform. Working closely with Lilt we’ve been enabling the first pass translation to be handled by AI, massively decreasing the time to moderation.
To further refine the result, we developed a Labs mobile feature allowing users to opt into Beta languages as they are translated. This provided direct in-context feedback on the exact words and phrases that required improvement.
The Impact
This collective flow promises to set a new standard for translation speed, quality, visual accuracy, user feedback and full publishing of new languages across the whole digital platform.