TRANSLATION IN THE 21st CENTURY
Translation as a utility
For a long time entrepreneurs in the translation industry complained that their products and services were never really appreciated by their corporate customers. Translation was an afterthought, something a secretary would worry about when she realized that the addressee did not speak ‘our language’. Is this still true today? How relevant is translation to corporate strategy? Do C-level executives really worry about language(s)? Why spend money on translation? Is it simply the cost of doing business, or is it the key to new markets? Does translation help save costs or generate revenue?
Download the PDF (180 Kb) and read later
In defense of a translation budget of roughly a billion Euros per year, the European Commission Directorate General of Translation (DGT) - the largest translation buyer in the world - has calculated that translation costs each citizen in the European Union just €2 a year. Surprisingly, this figure corresponds globally to what market researchers believe to be the size of the worldwide translation industry: around €15 billion. But even more interesting is that translation in this perspective is conceived as a utility paid for through taxes. While the business world is still struggling to justify the cost of translation, a new vision is gradually emerging: translation is part of the solution, not the problem - translation as a standard feature, a ubiquitous service, either free or paid for in a variety of ways appropriate to context.
We envision a 21st century world in which translation of any information into or from any language is always instantly available - paid for through taxes or as a bundled utility service. According to representatives of the DGT, even this most ambitious of services translates only a fraction of what needs to be translated currently. We estimate that the total cost of approaching the vision of translating everything on-demand could add up to an average of €10 per world citizen or €70 billion. We believe there will be many parties interested in subsidizing this spend, for the political, economic, social, cultural and security benefits it confers.
In a world with increasing volumes of content from and for everyone we cannot expect that the traditional publishers and enterprises will continue to pay the translation bills on the same basis as in the past. Translation would quickly become too expensive for any single organization (public or private) to carry entirely on its own budget. The solution demands the perspective of shared services. Translation in the 21st century will be a basic utility for everyone on the planet - a human right that everyone can demand and expect. Electricity, water, roads, the Internet, and language translation are all part of the basic services that help drive civilization as we know it.
Where does that leave translation as a business? Companies will move away from tactical, after-the-fact product localization to a more proactive enterprise-wide language strategy. They will tap into shared utility services with access to instant (if less than perfect) basic translation. But to differentiate themselves and win high-value customers, there will always be a critical need for high value-add personalization and localization services. Translation as a utility will have a profound impact on professional translation methods and approaches, enabled by a new generation of technologies embedded in enterprise information systems.
If this vision materializes, translation will become a business-strategic activity for the first time, and our translation models, ideas, and approaches will have to change. When translation is the solution, it plays a new and different role in information and communication systems.
Translation technology moves ahead
It doesn’t really matter whether technology is the cause, or the consequence, of this far-reaching vision of a world where language barriers are fading. Technological advances and business innovation are intertwined, and are now moving forward at an accelerated pace. Think of the impact of connectivity, and the amazing fact that the populations of online social communities are starting to outnumber some of the larger nations. Think of the potential power of these communities when it comes to resources for translation, customization and editing. Think of translations as data and how these data – in volumes of millions and billions of words – can help train new machine translation engines. Now even the most hardened skeptics are often surprised to see the quality of machine translation output.
Only five years ago people were pretty dubious about TAUS’ predictions about the rapid growth of automated translation.
More than half a century of research in translation technology has laid a tremendous foundation for further improvements in the 21st century. And obviously it is not just a matter of feeding more data to the machines. The research community is working on new generations of MT technology that explore the application of linguistic data such as syntactic categories and semantic analyses to the widely used data-driven tools and engines. There are signs that new hybrid MT engines, combining a variety of technical approaches tuned to the task in hand, will outperform current generation models in the near future. These techniques will help some languages more than others, but the refinement of technologies and gradual evolutionary improvement will bring much better MT for all language pairs.
Will there be a fundamental scientific breakthrough in language translation? Probably not. Will there be an incremental, and accelerating, improvement in translation technologies? Absolutely! In the absence of a scientific paradigm shift, solutions will piggy-back on the collective human knowledge of language. We are inventing smarter ways to do online collaboration and to attract and engage larger communities who contribute to the collective language knowledge. Solving translation problems can be as entertaining as online games.
IBM engages its own global workforce in making the IBM translation engine perform better. And while Google keeps pushing the frontier with more and more crude data, it discovers the limits of what is possible; automatic translation for a record number of 100,000 language pairs generates stupid mistakes. In a recent interview in Forbes Magazine Franz Och (Head of the Google Translate Division) says: “The trajectory we are on just doesn’t seem likely to reach artificial intelligence.” Computers need to learn how to learn things, like the human brains that feed them. Google’s invitation to its visitors to make better suggestions for stupid translations is just another step to help us get there.
Translation technology has made a generational leap forward in the last five years; the next five will see expansion and refinement of tools that support ubiquitous, utility translation.
