Reduced complexity of a powerful machine-learning model could
help level the playing field for automatic voice recognition all throughout the world.
With the emergence of virtual assistants like Siri, automated speech-recognition technology
has become more ubiquitous, yet many of these systems only work well with the most widely spoken
of the world's nearly 7,000 languages.
Now we'll talk about the new developments that machine learning has brought to auto-mated voice recognition.
Because these systems are mostly absent for less prevalent languages, millions of individuals who speak them are blocked off from a wide range of speech-based technology, including smart home devices, assistive technologies, and translation services.
Machine learning models that can learn the world's uncommon languages, which lack the enormous volume of transcribed speech required to train algorithms, have recently become possible thanks to recent breakthroughs. However, these methods are frequently too complicated and costly to be broadly used.
Researchers at MIT and other institutions have now devised a simple strategy for reducing the complexity of a sophisticated speech-learning model, allowing it to run more effectively.
www.dprg.co.in
