• February 2, 2015
  • By Leonard Klie Editor, Speech Technology and CRM magazines
  • FYI

Skype Translator Emerges

Article Featured Image

Microsoft in mid-December released a preview version of its long-awaited Skype Translator, a speech-enabled add-on to its popular Internet-based video- and voice-calling software.

Skype Translator has been in development for more than a decade. It features speech and machine intelligence technologies that convert spoken words into text, translate that text, and then synthesize it back into spoken words in the language spoken by the person on the other end of the call. The translations are also displayed on users' screens in real time during the conversation.

The software uses machine learning and neural networking to improve recognition and translation accuracy over time. The training data for improving the speech recognition comes from translated Web pages, videos with captions, and the transcripts of previously translated conversations.

This training data is what separates Microsoft's technology from competitors in the automatic translation space, Bill Meisel, president of TMA Associates, explains. "Because Microsoft uses a statistical approach to speech recognition and translation, the system can learn from experience and should continually get better as it has more data to analyze," he says.

But that also presents challenges. Skype Translator, Meisel says, "should be excellent for typical consumer conversations when the participants are talking about subjects that others have talked about. It may be more questionable when the topics are specialized, where there may have been few relevant examples for the machine learning to process."

Skype Translator is also limited in that the current version only translates spoken conversations between English and Spanish. The software can, however, translate text-based instant messages between 40 languages.

Microsoft was already using some of the translation technology underpinning Skype Translator to power its Bing Translator service and translate its product manuals and hundreds of thousands of support documents into languages other than English.

And while the technology is far from perfect, it has captured the attention of others in the speech translation market. "All of us in the international speech translation arena are delighted with every step taken toward global language comprehension," says Sue Ellen Reager, CEO of International Services and president of Translate Your World, which has developed its own translation software. "The world is ready to internationalize business, education, and conversation, for which affordable language speech translation is mandatory. Skype Translator is one of the steps toward this goal."

Translate Your World in early January launched Version 2 of its own machine translation software, which has many of the same capabilities as Skype Translator, with a few notable differences. For one, it supports 25 languages.

Translate Your World 2 also is a considerable upgrade from the previous version, which was strictly a one-to-many translation tool for Webinars and conferences. With Version 2, almost anyone can talk and be understood. Whether on Web conferencing technologies (such as Skype, Webex, Adobe Connect, Google Hangout) or tablets and smartphones, both sides of the conversation are translated. Moreover, people of many languages can speak together at the same time, all in their native tongues, with subtitles flowing in the language of each participant.

Translate Your World Version 2 also brings together speech and translation software from Microsoft, Google, Nuance Communications, SDL, ReadSpeaker, SIMS, Yandex, Systran, Android, KeDaXunFei, Baidu, and others.

For its part, Microsoft preceded the Skype Translator launch with a private beta trial version a month earlier. Microsoft acquired Skype, a Voice over Internet Protocol provider, in May 2011 for $8.5 billion. It was Microsoft's biggest purchase ever, and perhaps the riskiest, according to many analysts.

But Gurdeep Pall, corporate vice president of Skype, says the investment is finally starting to pay off. "Skype Translator is a great example of the benefit of Microsoft's investment in research," he wrote in a recent blog post. "We've invested in speech recognition, automatic translation, and machine learning technologies for more than a decade, and now they're emerging as important components in this more personal computing era."

Regarding the future of the technology, Pall also stated in the blog post, "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 platform for our more than 300 million connected users."

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues
Related Articles

Skype Translator Trials to Start Soon

Microsoft is now taking pre-registrations for Skype Translator.

WebRTC Enables Dual-Language Speech

Anticipating the powerful impact of cross-language talking.

Translate Your World Launches Medical Translation App

The hospital app can translate speech in up to 78 languages.