Microsoft released a preview of its Skype Translator technology today as an app for Windows 8.1. Skype Translator, first demoed in May during the Code Conference in Rancho Palos Verdes, Calif., provides near-real-time translation of typed and spoken messages.
At the event, Gurdeep Pall, corporate vice president of Skype and Lync, carried on a Skype video conversation with a German colleague, Diana Heinrichs. Skype Translator wowed attendees with its ability to provide on-the-fly spoken and text translations. A few hiccups aside, the app and its underlying cloud-based technology generated accurate translations and kept up with the speakers.
Today, the general public can take a crack at it, but with some limitations.
Pall returned to the Skype Big Blog on Dec. 15 “to announce the first phase of the Skype Translator preview program.” The program covers just two spoken languages, English and Spanish, for the time being.
Fans of Skype’s instant messaging capabilities have more options. Pall revealed that “40+ instant messaging languages will be available to Skype customers who have signed-up via the Skype Translator sign-up page and are using Windows 8.1 on the desktop or device.”
Microsoft envisions that Skype Translator, and technologies like it, will enable information sharing and collaboration in personal, education and corporate contexts that transcend borders and languages. “Skype Translator will open up endless possibilities for people around the world to connect, communicate and collaborate; people will no longer be hindered by geography and language,” said Pall.
Machine learning makes Skype Translator tick, Microsoft Program Manager Mo Ladha and Chris Wendt, Microsoft Research group program manager, explained in a separate Skype Garage post.
“Skype Translator’s machine learning protocols train and optimize speech recognition (SR) and automatic machine translation (MT) tasks, acting as the glue that holds these elements together,” they explained. “This ‘glue’ transforms the recognized text to facilitate translation.”
Error rates were reduced with the “advent of Deep Neural Networks (DNNs) for speech recognition, pioneered by Microsoft Research,” said Ladha and Wendt, enabling the software giant to begin commercializing the technology. “At the same time, the dream of global human-to-human communication was a major motivating factor and driving force for the MSR [Microsoft Research] researchers working on this technology,” they added.
The more users who put Skype Translator through its paces, the better it gets, said Pall. “We are starting with English and Spanish, and as more people use the Skype Translator preview with these languages, the quality will continually improve.”
Encouraging users to sign up for the preview, Pall said Microsoft needs the public’s “help to expedite new language releases.” Ultimately, the company hopes to blanket the globe with the tech.
“Our long-term goal for speech translation is to translate as many languages as possible on as many platforms as possible and deliver the best Skype Translator experience on each individual platform for our more than 300 million connected users,” said Pall.