Skype, a Microsoft acquirement as of 2011, released a Star Trek-esque universal translator at the Re/Code Conference kick-off this week. Skype Translator, as it is referred, will be a real-time translator tool that enables users previously barred by a language barrier to instantly communicate with one another. Microsoft explained the technology in terms of its synergistic innovations. Skype Translator, now in its beta phase, is the result of over a decade’s worth of study that exemplifies Microsoft’s bid to invest in long-term research. According to Gurdeep Singh Pall, the corporate vice president, the translation tool can go gold with the help of Skype users. The catch: Skype wants to use users’ conversations to help master a given language.
Before the privacy conspiracies get underway, the reason Skype would like to use users’ conversations is because the translator is designed to apply universally, that is, it seeks to function on a level beyond the restrictions of formal speech. Understandably, informal speech is liable to foster a wealth of translation errors, and that goes for inter and intra language communication. Lingo can be diverse even within the context of a single language, so it is not difficult to anticipate the barriers that may arise when more and more languages are added to the program. As of now, Skype plans to launch the translator tool sometime this year. The tool will come equipped to translate some 40 languages in real time, with the intention to add more.
According to CEO Satya Nadella, 15 years ago Microsoft began research on “speech recognition, machine translation, and speech synthesis.” At the Re/Code conference, he described Skype Translator as a “deep neural net” that synthesizes a model capable of speech recognition. Nadella elaborated further on the capabilities of the model, explaining how it has shown the capacity to “learn.” With each new language added to its repertoire, the model has shown itself to be able to learn more about previous languages in the databank. The ability of the model to do this remains a mystery to Nadella and the developers, who remain in awe of “its brain-like” capacity to learn.
For people familiar with Star Trek, this technology is recognized as the “universal translator” technology. In the series Star Trek: Enterprise, the universal translator was introduced as a prototype. The translator itself was a handheld device that could be attached to a communicator, a glorified walkie-talkie much like the one used by Captain Kirk in the two recent J.J. Abrams adaptations of Star Trek. Later development of the technology enabled crewmembers to attach a smaller universal translator to their uniforms. Most ships had communication systems equipped with the universal translator to help facilitate interactions with new species.
Interestingly, the purely conceptual and imagined technology of Star Trek’s universal translator anticipated some of the bridges that the Skype Translator has had to cross. Both translator technologies depend on exposure to spoken language in order to master it effectively. It is not simply about generating text-to-text translation. Any person who has tried to get out of Spanish homework can testify to the shortcomings of literal translation tools. Much in the same way a Spanish professor explains to her students that traveling to Mexico and interacting with native speakers will improve communication skills far better than simple rote, the Skype translator has shown a mysterious capacity to improve with more exposure to spoken language.
What remains fascinating about the Skype Translator is this ability to learn from exposure. Any regular translation service, like Bing, provides text-to-text translation based on preexisting language parameters. Microsoft’s innovation is to include a voice over, as well as text-to-text translation. For this reason, the mastery of spoken language is key. In order to optimize exposure to real life spoken language, Skype hopes that its users will help by allowing developers to monitor conversations. However, the company has yet to elaborate on how that procedure would be carried out. Until then, the privacy-related conspiracy theories may be allowed to fester.
By Courtney Anderson