As a teenage boy, I remember reinventing the wheel a couple of times. One of these “reinventions” was translation memory technology. As an aspiring translator back in the 1990s, I was dreaming of computer software that would recycle translated segments and recognize translatable words from a digital dictionary. Only years later, as a university student, did I realize that my invention already existed. To my great disappointment, Trados was widely used and its technology was very well developed. Although adaptive machine translation (MT), in-context exact matches and predictive typing were still to come, developments in the 1980s and 1990s opened up the door for a century characterized by technological inventions — and the translation industry was not an exception. It was still a time of transition, the early days of a new era: Google Translate was not yet conceived and when typing CAT tools in a search engine, it asked whether you meant CAD (computer-aided design) tools.
Our industry has been doing a pretty bad job in objectively educating newcomers about existing tools and technologies. To start with, books, reference material and eLearning courses on the subject are scarce — for example, there’s nothing to be found on Coursera. What’s more, only a few universities apply translation technology to their teaching in the translation classroom. And the attendance at industry events of translation students, newbies and startups is negligible.
The current TAUS Translation Technology Landscape Report published in October 2016 is a total remake of a previous edition with the same title published by TAUS in 2013. That previous, 70-page work covering the global translation technology sector has now been completely updated and extended. The raison d’être of this latest edition is the emergence of several new trends and technologies. Yet the report is not complete, and of course it cannot be. The landscape is ever-changing with many technology startups coming and going. So, a number of tools are missing from this report and it is probably better this way. The technologies that count are there.
The latest discussions around neural machine translation (NMT) and the negative buzz Google received for its “bold statements” on bridging a certain gap show that some of us are still rather skeptical about new findings and inventions in translation technology today. Still and precisely because of the new hype around NMT more and more companies (language service providers mainly but also some buyers) are now considering the possibility of giving machine translation a second try. This is good news for translation technology providers, for consultants, hopefully also for translators and, after all, good news for the whole industry.
The first section of the report is a historical overview of the translation technology sector, exploring the changes and the challenges of the industry in a six-decade overview.
The next section offers a more in-depth analysis of the sector offering a categorization of tools and technologies. It is interesting to see that in our current decade already, more new technologies have been released than in the three previous decades altogether, and that Europe is leading the technology innovation in our industry with more than half of the tools described in the report. The 82 tool profiles contain well-established solutions versus the latest technologies: machine translation technology, quality assurance tools, translation memory tools, globalization and translation management systems and so on.
Finally, the trend section deals with the current technological trends that are expected to play a major role in the evolution of the translation industry. Some trends, such as NMT, have just left the restricted circles of academic research and are entering the commercial arena while others, like cloud, have been underway for some time now and have reached the plateau of productivity.
Trends
The section on trends can also be viewed as a crash-course on the five main trends that will dominate the years to come: 1. The cloud will become pervasive 2. Translation gets datafied 3. MT marries with artificial intelligence (AI) 4. Translation quality becomes quantified and benchmarked and 5. Speech-to-speech translation will become an alternative to text translation.
In all the trends described in the report, data plays an important role. We are becoming more and more data-driven. There are many data points to be harvested and analyzed in various stages of the translation process. The 82 translation technologies profiled in the report potentially record time spent on segments, keystrokes, edits, error annotations, time stamps and so on — all the information needed to create meaningful reports. Still, the translation industry is lagging behind when it comes to making intelligent use of the collected data. Other industries (such as the travel industry) made the switch to data-driven decision making a decade ago. In our industry, using dashboards to monitor every single data point in the translation supply chain is a privilege of the few. Apart from some large enterprises, most buyers and vendors of translation services still use intuition or rely on very limited data sets to make business decisions.
The questions that occur now are how to aggregate the collected data, how to interpret the results and how to improve and even try to automate various processes based on this data. This new trend unavoidably leads to dashboards, benchmarking and machine learning.
Another trend is MT empowered by new methods originating in AI. Today, more than 99% of all translations are produced by MT engines. MT is invisible but at the same time it is everywhere. To the question of whether or not this means the end of the translation profession, the report answers with “not yet.” Although futurologists like Ray Kurzweil predict human levels of translation quality by the year 2029, the computer scientists TAUS interviewed could not confirm this claim. Despite the amazing pace of technology development, true mastery of language is going to remain very much out of reach for software, at least in the near future. According to the report, the role of professional translators will not vanish, but it will evolve through technology.
In the section on cloud technology we read that more and more businesses are migrating to the cloud. To meet their needs, a growing number of cloud-based solutions are being built, producing high value, vertically focused solutions. The report estimates that 90% of new software applications will be deployed in the cloud.
The TAUS report also predicts that the current spree of startups in translation technology is only the beginning of what we can expect when convergence comes to full maturity. Translation will become available on every screen, in every app and on every signboard in the coming years.
In previous decades translation technology was concentrated on productivity gains in the area of user documentation. In the convergence era we will see a rapid expansion of translation technology categories with, for instance, proxy-based translation platforms for easy translation of websites, dedicated tools for localization of mobile apps, tools for translation of subtitling and community-trans-lation platforms for translation of user-generated content.
