How AI is Bridging the "Communication Gap"

For people who suffer from motor disabilities, the task of communication can be one that plagues their everyday. Even with the aid of modern technology, when relying on computers to speak, the rate of communication is typically around 80 to 135 words per minute less than the average person. For something most take for granted, you might not realise what an effect this can have on someone's day to day interactions and quality of life.

The use of computers to act as people's voices is nothing new, but the technology has been still wrought with errors, and too slow to make the conversation flow at a natural rate. Our voice is a powerful tool, and unique to the individual, so how is AI being used to predict the specific speech patterns of a person and finally bridge this gap?

 
credit : techcrunch

credit : techcrunch

 

Research shows that, as is for others, the reuse of the same sentences and phrases is a common trend for those using speech synthesis technology. While inputting these common phrases was a time consuming and inefficient task before, we can now better predict the desired text using AI that has analyzed the tendencies of its subject.

A recent study by the University of Cambridge and the University of Dundee made further strides in improving the access to speech for non-verbal people, by incorporating contextual elements to assist in speech prediction. Now even location, time, and the identity of the person being spoken to can become factors that influence the decisions of the assistive AI and with great results; their new method using these "context clues" was shown to reduce the number of keystrokes needed to communicate between 50% and 96%.

The first study of its kind, utilising the context of the speaker hand-in-hand with speech synthesis technology, shows exactly how instrumental AI can be to improving those affected by motor disabilities.

Further reading - https://www.cam.ac.uk/research/news/ai-reduces-communication-gap-for-nonverbal-people-by-as-much-as-half

Thumbnail credit: https://medium.com/@RAF100STEAM/6-problems-artificial-intelligence-faces-in-speech-recognition-ae705cfa1a72