Microsoft Demos a Translator That Re-Creates the User's Own Voice

Article Featured Image

Microsoft has unveiled a speech-based translator using text-to-speech technology that enables users to speak 26 languages in their own voices.

In a March blog post, the company said that by cross-training a text-to-speech system, the translator can correctly pronounce words from multiple languages, even those built from the voice samples of someone who only speaks one language.

"Out of a speaker's monolingual recordings, our algorithm can render speech sentences of different languages for building mixed-coded, bilingual TTS systems," Microsoft said. "We have recordings of twenty-six languages, which are used to build our TTS of corresponding languages. By using the new approach, we can synthesize any mixed language pair out of the languages.

"Sometimes it is even more desirable to have the response in mixed languages," the post continued. (For example, in a foreign country, it would be convenient if a user of a car-navigation system who is not fluent in that particular foreign language can hear instructions in mixed codes. Entities like street names would be synthesized in the local language and routing directions would be in the user's native language.) "The mixed-coded TTS can be easily built by a truly bilingual speaker," Microsoft explained.

The company also demonstrated another language translation project, Microsoft Translator Hub, which enables businesses and communities to build, train, and deploy customized automatic language translation systems between virtually any pair of languages. Using Windows Azure, Microsoft Translator Hub is an extension of the Microsoft Translator platform and service. The more data users input into Microsoft Translator Hub, the more it learns and the richer the insights and results for future translations, according to the company.

"Until now, automatic language translation has been at the behest of a few major providers," a company spokesperson said. "As a result, the number of covered languages is limited to less than 100, compared to the more than 7,000 languages currently in use around the world. This both limits the native speakers' access to the global knowledge base and puts increasing pressure on the adoption of nonindigenous languages simply because that is what is most readily available with today's technology. Microsoft Translator Hub enables the development of custom language translation between virtually any pair of languages via cloud and Web services."

Microsoft said organizations can build customized translations that are scoped to their offerings and terminology by combining some of their own existing translation data with the Big Data from Microsoft's Bing search engine. The new systems are then deployed into a private space, where invited members and reviewers can help refine the system's accuracy.

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues