Skype has started to roll out the "first phase" of its live speech translating feature. It is between Spanish and English for now and there is also text translation for instant messaging in more than forty languages.

As indicated back in May, see Skype Translator - Communication Without Language Barriers the initial offering is for Window 8.1 and also for the Windows 10 Technical Preview. 

Although it is currently limited to English-Spanish it is an important milestone for Microsoft Research which has been a pioneer in language translation. 

According to the Skype Garage blog:

Skype Translator is the result of decades of research in speech recognition, automated translation, and general machine learning technologies, combined with an intense focus on the user experience. 

Microsoft was one of the first to delve into the challenge of speech translation. Recent improvements in speech recognition, made possible by the introduction of deep neural networks combined with Microsoft’s proven statistical machine translation technology, allow for better translation outcomes, making meaningful one-on-one conversation possible. 

Click for high resolution pdf (about 1Mbyte)

As the infographic shows machine learning is used in various ways during the live translation process:

• Speech Recognition -  compares speech with audio snippets from millions of previously recorded samples and transforms the audio to a set of text candidates
  
  
• Speech Correction - detects and removes the "ums" and "ahs" together with stutters and repetitions and makes the best selection among any sound-alike words 
  
  
• Translation - reviews millions of pieces of previously translated content to provide a translation that "aligns" one language with the other before using text to speech to produce the audio result.

The infographic also indicates how use by real users should improve the system. I imagine that most 21st century grandmas would be a bit non-plussed by the use of "Usted" in the context of a chat with their granddaughters! 

This video shows how delighted some of the early beta testers were with the software. Students from two elementary schools, one in Mexico, the other in the US, engaged in a Spanish and English game of ‘Mystery Skype’ in which the children ask questions to determine the location of the other school. 

Skype translator is a much needed tool and as it is used the machine learning process should ensure that it improves until we really do have the longed-for Babel Fish that we simply plug into our ears.

Comments

Make a Comment or View Existing Comments Using Disqus

or email your comment to: comments@i-programmer.info