Dragon speech recognition turns 20

Macworld Australia Staff
1 September, 2017
View more articles fromthe author

Nuance Communications has announced that its signature software solution, Dragon, turned 20 this year.

In the early 1980s technology experts were struggling to develop a system that would understand about 35 words that were spoken carefully, deliberately and in isolation. But today, Nuance’s Dragon can understand natural dictation from pretty much any device and from any reasonable environment. Since those early days Dragon has improved its average error rate by around 18 percent each year.

“The vision from the beginning was that Dragon would transform the way people create documents,” says Derek Austin, Dragon business manager, Asia Pacific, Nuance. “Instead of typing contracts, reports, correspondence, records, case notes and so forth, people would use their voice to create documents to help boost productivity.”

These improvements have occurred thanks to a combination of factors. Algorithms applied to recognising speech have been getting steadily better, while more powerful computing platforms allow Nuance to use more training data to improve its models. However, the most transformative factor has been the influx of deep learning into speech recognition.

So what does the future hold for Dragon? Its versatility is key. Not only will the software continue to drive document creation, its speech recognition capacity underpins the capabilities that virtual assistants will deliver in the future.

“As virtual assistants evolve, they will probably follow the user from device to device, and a combination of natural language understanding, artificial intelligence, voice recognition, knowledge representation and other modalities will allow the user to interact with an immensely rich world of content, services and smart devices,” Austin says.

Leave a Comment

Please keep your comments friendly on the topic.

Contact us