TAUS is starting the new year with something special - a new eBook! We warmly invite you to download and read Nunc est Tempus, activate your (biological!) neural networks, and join the conversation about the future of the industry.
Why now, you might ask? We attempt to show that the stars are aligning on a new paradigm for the translation service industry. Technology, datafication, and the global economic context require a truly 21st century model of translation practices. So we have interviewed a number of leading visionaries from the industry and marshalled their insights into a story with yes, a revolutionary message.
To give you a flavor of what we propose, here are three drivers of change we identify in the book – the featurization of translation, the impact of machine learning, and how translators are set to play a new role in the industry.
Translation becomes a Platform Feature
Like most industries, translation services follow a sigmoid growth curve – the S-curve. By constantly adding intelligence, an industrial process can rapidly scale up, expand and then level off as it compacts into a feature in some larger process. Translation began in the pre-digital age as a scattered collection of suppliers using a manpower-intensive process to deliver services. It then gradually scaled up and structured itself into large multi-service companies that applied technology fixes to optimize production.
Today, as it nears the top of the S-curve, it has reached a developmental plateau: translation as a process has now become compacted into a feature on various digital platforms. In the extreme case, translation services can be found packaged as a feature on digital watches, earphones and smartphones – the new consumer platforms. TAUS has been calling this recent “utility” stage of development as “translation out of the wall.”
This doesn’t, of course, mean the end of translation as a rich business service. But the translation industry must now adapt to a more streamlined delivery model in order to survive in a world where technology intelligence is replacing older methods. It needs to offer new voice and text-based services, take on board new language pairs and content types, and solve new data problems. Above all, it must address the core issues of workflow, quality, cost and delivery time in new ways to handle the challenges of “digital transformation.“ In the book, we propose a powerful workflow model that factors datafication, quality evaluation and tech-driven processes into a dynamic whole.
The Teachings of Machine Learning (ML)
Just as the S-curve takes us from dedicated translation organization to becoming a feature in other platforms, so technology tends to package, automate, miniaturize and connect. While the first machine translation systems were stand-alone mammoths, today’s machine learning neural MT (NMT) solutions (software that learns translation patterns from data) operate as part of a global community of research and development focused on reiterative learning.
At the same time, it is important not to confuse the growing NMT trend with the promise or hype of other AI solutions, as in healthcare or electric vehicles. Translation tech is not making predictions or decisions based on data that determine critical outcomes. So far it simply accelerates the transfer process to make a new language version available for improvement where necessary. It is equally likely that NMT will not be the last word in translation automation, even if its time has definitely come today. What we emphasize in the book is that we must learn to anticipate – even demand – more improvements to the technology at a faster pace in the coming years. Translation businesses, along with translators, will need to grow more agile to respond rapidly to shifts in the compute-translate-communicate framework that now molds our practices and decisions. This is why we insist on learning how to ride the S-curve, and not struggle against it. Creative reinvention will be the name of the game for the industry in the foreseeable future.
For too long, the industry has allowed outsiders to spread the “Is MT as good as human quality translation?” meme. We all know the answer, which is usually irrelevant to serious business and technology decisions. Now is the time to move on to more interesting debates. MT clearly delivers translation at scale faster than humans; humans clearly ensure appropriate quality levels more efficiently than machines.
Translators will continue to deploy a growing array of technology aids, and LSPs will continue to explore the market for translation automation. As we show in the book, the market is opening up new opportunities for language-specific tasks beyond the sole medium of text. Multimedia will increasingly require a richer, multi-channel, multi-skill combination of talents that can be tailored to tasks old and new through a mix of data science and human intelligence.
There is always a need for translators who work directly with clients to craft eloquent translations for well-known use cases, and a need for large-scale automation to handle massive multilingual content reproduction, as well as every use case in between. So now is the time to focus on how metrics and data can support an optimal translation and review process. And how innovative ideas in training and subject-matter expertise can help translator-reviewers expand their role in content localization.
We also examine numerous other interesting touchpoints in the translation journey, from Chinese-centrism to the challenge of ambient intelligence. For all this and more, download Nunc est Tempus here and join TAUS’ 2018 quest to build truly 21st century translation workflows.