Yesterday and tomorrow: Making the promise of technology a reality
TWB's Words of Relief project provides a one-of-a-kind language service to improve communication between crisis-affected communities and humanitarian responders, so that life-saving and life-changing information is not lost in translation. We train interpreters and translators in the languages of crisis-affected countries, crowdsource rapid-response teams of professional volunteer translators for urgent communications, and promote the use and development of machine translation and other language tech tools for emergency response.
Yesterday, 34 people drowned in the Mediterranean in a botched rescue attempt. Cholera took hold in Yemen, with hospitals filling up. In South Sudan, famine spread, beating war as the crisis-of-the-day. And in Bangkok, young Afghani men hoped against reality that, someday, they would at least step out of slavery.
One week, four stories. Despair spread across thousands of miles. And nothing new, in a sense: year on year, people die of preventable hunger and disease, or leave their homes praying for a better life that rarely materializes.
All too familiar, but not inevitable. Within Translators without Borders’ Journey to Scale, we are exploring the next breakthrough, focusing on making more information available in the right languages.
The commonality across these stories is lack of free access to trusted information. The reasons run the gamut, from inability to get health information in remote areas, to lack of education and high illiteracy rates, to minimal access to technology, to pure manipulation of information. Whatever the reason, those who do not receive information, or don’t control what information they do receive, lack power. They rely on the goodwill of others who have access to information, or the good fortune that someone they trust provides information in a way that enables them to help themselves.
What happens if we give that power to the billions who lack it? We don’t really know. But it is worth trying as we strive toward a world of reduced disease and poverty and greater equality.
In my world at Translators without Borders, that means using language technology to unlock vaults of information available in ‘major commercial languages’. Technology is dramatically improving machine translation (MT) among most European languages, and even into Arabic from French, English and a few more major languages. The relatively new field of neural MT holds great promise.
In the words of Mark Zuckerberg, “Throughout human history, language has been a barrier to communication. It’s amazing we get to live in a time when technology can change that. Understanding someone’s language brings you closer to them, and I’m looking forward to making universal translation a reality.”
But that reality is very far off for billions of people whose languages are not ‘commercially viable.’ Neural MT requires massive ‘parallel datasets’ (identical texts in two languages) to train it; even statistical MT requires upwards of a million strings to create a viable translation engine. That data does not exist yet for most languages. But the promise of building it – enabling a Hausa or Somali speaker to translate information they want, in text or audio – is a game-changer in humanitarian response. True empowerment. With commitments from major tech developers like Facebook, Google and Microsoft, we can get there.
Language access to information opens up an entire world, not just for basic health and protection information, but for online education, advanced medical information in remote areas, and even cultural information that enriches lives. This is especially true for non-literate populations, who are disproportionately women. In a context of population displacement, it allows developers to create apps specifically for refugees in their own languages to help them access vital cash, food and labor opportunities. The promise is endless.
To get us there, Translators without Borders, as part of its commitment to explore new technologies for delivering its Words of Relief program, is developing the Common Language Initiative (CLI). The CLI signals a new approach to information access for crisis-affected people: ensure people can get any information they need and want in their language. By developing large datasets via TWB-trained translators and post-editing experts, by ensuring that the data is captured and, finally, put to use to build better MT, we can make translations of humanitarian content 50 percent faster without compromising on quality, thus significantly reducing costs. It has the added benefit of ensuring that, when people do have connectivity, they will be able to get ‘good enough’ translations of any information on the internet using a simple online translation tool.
The CLI is a major effort. It involves commitments from major technology companies which will use the datasets to develop MT in major ‘pivot’ languages (those, such as Hausa or Swahili, that are spoken as second languages by large populations, so an English-to-Hausa MT engine could enable rapid English-Kanuri translation, pivoting through Hausa). It also includes content providers who have very large volumes of content for parallel datasets, and international humanitarian organizations that will develop apps using the new MT engines.
The CLI is a global initiative with local communities at its center, as both translators and consumers of information in marginalised languages. TWB is drawing on its experience with training community language teams in Kenya, during the initial Words of Relief pilot, as well as further work and investigations in Greece and Nigeria, to develop a sustainable method for community development and ongoing involvement.
Information is the great equaliser. For those of us who have freedom of information, that means figuring out how to bring the same freedom to the rest of the world.