Every four years the Olympics bring the world together, albeit briefly. And no matter a person’s global location, we all connect through athletes like Oscar Pistorius, Jessica Ennis, or Usain Bolt. That’s because the water cooler has been replaced by social media and real-time global discussion. Still, language barriers prevent us from completely connecting. But what if the language barrier didn’t exist? What if we could completely overcome it in the next ten years?
Jonathan Lichtman, senior vice president of SAIC, believes it’s possible. His firm recently launched the first translator that integrates both machine translation and automated speech recognition into the same platform. We spoke to Lichtman about why the possibilities of this technology are exciting and what lies ahead for the sector.
1. Why is translation technology important?
We live in an era of rapid globalization, which is demonstrated by the growing demand for language services. Common Sense Advisory, an independent analyst firm that focuses on this area, estimates the demand for language services will grow 12% annually. Simply put, human translators do not have the capacity to meet this demand.
2. So how exactly does translation technology help meet this growing demand?
Translation technology solves this problem in two ways. First, it enables translation at a level where it does not need to involve a human. Secondly, we’ve seen translation technology increase human translator productivity by up to 400%.
Ultimately, translation is important because it facilitates multilingual communication and allows people from around the world to better understand one another culturally, economically and socially.
3. Part of this technology involves developing a new language pair. How long does it take to develop a new language pair, and what are the main barriers to automating the translation?
A language pair will usually only be developed if it is commercially viable, and the market tends to dictate which language pairs get developed first. For example, in the U.S., Spanish, Brazilian Portuguese, and Asian languages are the most in demand.
It can take as little as a month to develop a new language pair, but the time certainly varies. What makes our hybrid machine translation technology so unique is that we can develop a new language with much less data than either statistical or rules-based engines alone.
The two main approaches to machine translation are generally rule-based (RBMT) and statistical (SMT). RBMT uses manually programmed rules to translate one language to another. SMT involves the use of previously translated content to determine which words and phrases have the highest possibility of conveying the correct meaning. Both have advantages and disadvantages, but by combining the two approaches together in the same engine — a hybrid approach — language pairs can be developed faster and with less pre-translated “training” data.
That is an important leap in translation technology because it addresses two main barriers to the automation and accuracy of a language pair: not having enough translated data available and the constant evolving nature of language.
4. With this new approach, at what point can we completely overcome the language barrier?
This is an interesting question that really shows the complexity of human language. If I am sitting in the same room as a colleague and we are both native English speakers, I still may only understand 98% of his intended meaning. So for that reason alone, translation technology will never be perfect. The rapidly evolving nature of language I mentioned earlier is another.
However, I do believe that there are currently tailored language solutions that can now approximate the accuracy of humans in some cases. I also believe that in the next decade the quality of language technology will continue to evolve where human-quality machine translations will be the norm rather than the exception. That said, human communication will never be without misunderstanding and an inaccurate perception regardless of how well the technology can perform.
5. As we move closer to better translation capabilities, what kind of impact will this have on cultures around the world?
I cannot emphasize enough the impact this technology will have on culture. Language and culture are inseparable and language plays a huge role in the transmission of different cultures across the globe. I see two significant benefits from the diffusion of culture through the use of this technology. The first is that different cultural perspectives will become available across a much larger audience. That will lead to better understanding of perspectives, even if it doesn’t mean that there will be agreement on all issues. The second is that the cultural diffusion will become much more accessible regardless of the socioeconomic conditions that may have prevented some cultures from either sharing their culture or taking advantage of new cultural influences that can now become available.
6. Who would you say benefits most from this technology?
Currently, larger corporations with business operations that span borders are seeing the most benefit because they are taking advantage of the technology to more effectively communicate both internally and with their customers. But as cloud and SaaS offerings are becoming more prevalent, per-word pricing has given mid-sized and small businesses access to corporate level translation software and leveled the playing field. This has driven growth by providing access to new customer markets and audiences.
Translation technology is just getting to the point where the general public will start to see its impact in their everyday lives. Not only are there free, general translation engines online, but even the more tailored, advanced technology is becoming accessible via mobile devices.
7. Is there a country leading the way?
It is a very difficult question to answer. There is no question that the United States has led in the development of both Automatic Speech Recognition and Machine Translation products. However, that development, and the progress that these technologies have made are a result of important research and development across the world. The global nature of these technologies cannot be overstated, and I believe that the future will hold even greater collaboration across the world.
8. So what kind of shift can we expect in this space moving forward?
Right now a really exciting innovation is the integration of machine translation (MT) and automatic speech recognition (ASR) technology (Apple’s SIRI is an example of ASR) into one platform. This allows for speech-to-speech and speech-to-text translation in near real time. By housing the two technologies on one platform it also makes it easier to localize into a smartphone, tablet, or other mobile device. In a way, it’s reminiscent of the Star Trek “universal translator.” This technology is here today.
Moving forward, translation technology will incorporate additional information sources from cameras or GPS technology to help improve the accuracy of translation. For example, if I say “meet me at the bank,” where you go would depend on whether I have a checkbook or fishing pole in my hand.
9. Can you give me some examples of how this tech is currently used to help make a given field better?
In the healthcare community translation technology can help alleviate part of the administrative burden of patient intake, enhance healthcare performance, and most importantly, increase patient safety, and improve health outcomes.
Another example is within the retail industry. Not only is the technology being used for external communication with customers who speak a different language, it is also being used internally to streamline the supply chain.
Image courtesy of iStockphoto, atakan