Rob - Fotolia
Talking to computers may be in the realms of science fiction, but as a technology it has reached the plateau of productivity in Gartner’s Hype Cycle of smart machines.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Microsoft’s Cortana, the iPhone’s Siri and Now from Google allow tablet and smartphone users to talk to their devices, but some industry watchers claim speech recognition has limited appeal for wide-scale adoption.
The analyst firm stated that speech recognition is a type of technology that has so far had limited applicability and adoption, but that could change very quickly.
“If it were to pass a threshold of capability similar to how humans recognise speech, we would see an exponential increase in adoption and possible areas of application. Recent advances in machine learning and the hardware involved are getting us close to this threshold,” wrote Gartner research director Magnus Revang.
Application areas include dictation, virtual assistants, browsing and menu navigation across PC and mobile platforms, as well as in-car computers.
But the technology also has a stigma associated with it. “Talking to devices is a good way to get yourself locked up,” said Gartner fellow Steve Prentice.
If the breakthrough does not occur, Prentice said the technology could become obsolete, or at least remain a niche. He cited the once ubiquitous mouse as an example: “The mouse is now superfluous. I have a mouse on my desk, it still makes sense to me, but as a mainstream product it is no longer relevant. Speech recognition technology is a mature technology, but it still has a long way to go.”
From an IT strategy perspective, Gartner warned that speech recognition is still very susceptible to the system’s immediate surroundings – environmental noise and distance between the user and the microphone dramatically affect performance.
Another problem Gartner identified is that cloud-based systems hamper response time, affecting transcription performance and, subsequently, adoption.
“Reliable speech recognition will remain elusive until machine-learning algorithms are further developed and processing tasks are more effectively partitioned between local processing and cloud resources,” said Revang.