Researchers from the universities of Cambridge and Dundee developed the system for people with motor disabilities, who often use computers with a speech output to communicate. Unfortunately, these tools are generally slow and error-prone. Research shows that people normally type between five and 20 words per minute, but speak between 100 to 140 words per minute. As a result, people who rely on computers to communicate can struggle to have meaningful conversations. [Read: Speech recognition technology is racist, study finds] The new AI tool helps fill this communication gap by reducing the number of keystrokes they need to communicate.
The importance of context
Everyone repeats certain sentences and phrases in conversation. However, existing speech synthesis technologies are slow to retrieve these sets of words. The new AI tool makes this process quicker. As a person types, algorithms analyze the text and the context of the conversation, such as the location and time of day. It also identifies the other speaker, through a computer vision algorithm trained to recognize human faces from a front-mounted camera. It uses this information to suggest sentences that are relevant to the conversation. The researchers say that this eliminates between 50% and 96% of the keystrokes the person has to type to communicate. “This method gives us hope for more innovative AI-infused systems to help people with motor disabilities to communicate in the future,” said Cambridge Professor Per Ola Kristensson, the study’s lead author. You can read the full research paper here.
