Can Microsoft Skype Translator Break the Human Language Barrier?
At the recent Code Conference, Microsoft showcased a demo of the Skype Translator app. The demo included Skype VP Gurdeep Pall speaking in English with a German-speaking employee and the employee responding in her native language, thus enabling cross-lingual communication.
Microsoft Research has invested a good deal of effort in perfecting the machine translation technology. Skype Translator takes the concept a step further by adding the capability to translate speech. As The Official Microsoft Blog explains, the key features of the demonstrated technology are speech recognition, machine translation, and machine learning technology.
Bing Translator, which makes use of Microsoft’s machine translation technology, already has many successful commercial applications. Twitter recently tested the Bing Translator technology for automatically translating tweets.
Microsoft CEO Satya Nadella had the following to say about the machine learning capabilities of the Skype Translator:
One fascinating feature of this is something called transfer learning and what happens is you teach it English, it learns English. Then you teach it Mandarin, it learns Mandarin but it becomes better at English. And then you teach it Spanish, it becomes good at Spanish but it becomes great at both Mandarin and English.
This is possibly due to Microsoft’s focus on research around neural networks, which imitate the way the brain behaves and allows machines to learn as humans do. Intense research led to the current levels of accuracy. However, in order for this technology to break the human language barriers, it must improve upon the current perceived limitations of the system.
The accent and delivery of speech varies from person to person and from culture to culture. This diversity makes it harder to form standard patterns to recognize and translate speech. Other factors—dependency on hardware, such as microphones and environmental noise—make it more challenging to deliver a flawless output.
As mentioned by Microsoft’s project leader Arul Menezes, "The technology is only as good as the data. One big focus has been to scale up the amount and kinds of data that go into the machine-learning training of these systems." The research team has explored various avenues, including the social media conversations toward this.
The success of this demo put a focus on language technology, but Microsoft is not the lone player in this segment. Last year, Facebook acquired Mobile Technologies, which is said to have done a good deal of research in the area of speech recognition and machine translation.
The world will look forward to the end of 2014 when Skype Translator will be available as a beta app. However, while acknowledging the positive buzz about the endless possibilities of changing the way we communicate, we should also remain pragmatic that the technology is still evolving and may take years to accurately simulate human speech.
Do you agree?
