Why TED Talks moved their global translator community to CaptionHub, and how we helped them do it

After an intensive deployment period, this week TED moved their entire TED Translator community to CaptionHub. To those who regularly watch TED events, it will come as no surprise that TED is hugely committed to making TED Talks accessible to audiences around the world.

TED has a highly engaged global community of 38,948 translators, translating into 115 languages. These volunteers, spread across 160 countries, subtitle 1200,000+ TED, TEDx and TED-Ed talks to appear and be accessible in 155 different countries. With around 3,600 hours of content to caption and translate each year, TED selected CaptionHub to improve their workflow and efficiency and to give their volunteer linguists the best possible user experience – wherever they’re located.

CaptionHub has been working closely with TED for two years, a period over which we became intimately familiar with TED's complex subtitling-at-scale requirement. On account of the necessary scale and required production standards, it was clear that TED had outgrown their community translation platform.

CaptionHub has always supported household brands and large enterprises in their subtitling workflow by providing cutting-edge technology. The scale and nature of the TED translator community, as well as the way in which that community workflow had evolved, meant we had to develop an entirely new approach in how we thought about subtitling workflow and scale.

Working with Jenny Zurawell and Helena Batt (Director and Deputy Director of Translation at TED, respectively) and their project and engineering teams, we embarked on an intensive multi-month engineering project. The outcome has delivered the most advanced subtitling platform available to the TED Translator community, whilst delivering important features for our existing enterprise customers.

We launched the initial tranche of workflow product work back in February which consolidated a streamlined approach for how large groups of linguists can use a self-serve workflow to choose video projects to translate, based on their language proficiencies. After a huge amount of complex and intensive product whiteboarding, envisioning and prototyping this work eventually led to a full re-architecture of our workflow model and associated user roles.

Teams who want their linguists to self-serve can choose from two approaches:

The Self Serve workflow is a new model of linguist assignment, designed for larger organisations with huge numbers of linguists or volunteers, but it works just as well for smaller teams who wish to work more flexibly. A key feature of Self Serve is that captions are viewable to linguists without them having been explicitly assigned.

The Self Serve with Review workflow builds on Self Serve. As the name suggests, Self Serve requires less intervention on the part of superusers or producers, allowing linguists to edit captions that they've chosen to work on themselves. When your workflow needs an additional layer of control and review after the linguist edit, the Self Serve with Review offers that capability. Once captions have been translated they're set to Ready for Review. Then reviewers are able to claim captions – in exactly the same way that linguists can. Once captions have been claimed, Reviewers are able to perform a variety of actions: they can edit, approve, sending the captions back to the original linguist, or reject changes entirely.

Both workflow capabilities have advanced controls such as claim expiry and assignment limitation.

When using Self Serve workflows, you may wish to restrict your users from claiming too many projects to work on at the same time. If that were to happen, then you may find that the linguist who claimed too many projects is overstretched, whilst other linguists have nothing to work on. Assignment Limiting lets you control how many projects any linguist can work on concurrently.

With Self Serve workflows, there's also danger that a linguist could claim a project to work on, and then abandon it. Unless a superuser or producer intervenes, then that project would effectively be locked to other users. Claim expiry helps avoid this by automatically unattaching linguists from a project after a specified time period.

Alongside a complex step-change in the product, TED and CaptionHub ultimately delivered a truly significant business transformation programme for TED and their global community. In a relatively short period of time, a technology change was designed, built and shipped alongside communications and project effort engaging a global community of 40,000 linguists. From our perspective it's been a real delight working so closely with such a dedicated, professional and excellent team at TED. We've extended CaptionHub's rich functionality to deliver a platform that will be at the very core of TED's mission to bring ideas to life for people of all languages.